I loved WordPress, I really did. I knew it in its very early days, when it was used to create simple blogs, without ready-made themes. WordPress solved a major problem back then: thanks to its back office, people who could not code were finally able to publish on the web easily.

For twenty years, the formula worked wonderfully. Even today, around 40% of websites run on WordPress according to W3Techs, a share that has started to decline. But watching the dazzling progress of artificial intelligence and modern architectures, my conclusion is clear: WordPress will gradually lose its position as leader. We have reached a point where the very concept of a CMS is outdated.

Is WordPress really dead?

No, not today. WordPress remains the most widely used system in the world and millions of sites continue to thrive on it. What is coming to an end is the model it embodies, that of a website you manage by hand from a dashboard. For a new showcase website, AI and a static site now do better, faster and for less.

What is a CMS, and why is it outdated?

A CMS like WordPress is a system that brings together in a single block the front end (the part the visitor sees), the back office (the admin dashboard), the MySQL database (where texts, pages and settings are stored) and PHP (the language that, on the server, builds each page from that data).

On every visit, the server queries the database, runs PHP code, loads the plugins and assembles the page on the fly. For a long time, this heaviness was the price to pay for offering a simple interface to non-developers. It no longer has any reason to exist now that artificial intelligence can work directly on the site’s files.

Why the back office is no longer needed

The back office was invented in the 2000s for a simple reason. It allowed someone who could not program to edit their website without touching the code, by filling in forms that the CMS then turned into pages.

That need has gone. Artificial intelligence now understands a request made in everyday language and translates it directly into file changes. No more need for forms, menus or a dashboard to publish an article, correct a text or add a page. You just ask, in a sentence like this one.

“Publish a new article, add a page to my website, adjust the offer’s prices by 10%, generate an animated vector diagram to illustrate the process and put the page online.”

The AI prepares the changes, generates the elements and puts everything online in a few seconds. This is especially true for a showcase website, whose content changes a few times a month and which has no payments or customer area to manage. There are still exceptions, such as an online shop or a site fed every day by a large team, but for the vast majority of business websites, the back office has become an unnecessary step.

7days.dev, my laboratory, proof in practice

To make this claim, I rely on concrete evidence. On 7days.dev, which I use as an experimental laboratory, I chose a modern architecture. The site is built with Astro, its content lives in a Git repository, and it is served by a simple server that only distributes files. I have described how and why I rebuilt it this way after its old WordPress version broke down. The benefits in practice are beyond comparison.

• Maximum speed, without effort. PageSpeed Insights gives 7days.dev a score of 100 in all four categories, on mobile as on desktop, with the main content displayed in one second. I did not have to install any caching system or optimisation plugin.
• An attack surface reduced to a minimum. According to Patchstack’s annual report, the WordPress ecosystem recorded 11,334 new vulnerabilities in 2025, 91% of them in plugins. On a static site, there is no accessible database and no PHP code executed on a visit, so most of these entry points disappear. The server still needs protecting, but you sleep much better.
• Vector visuals created by AI. Designing a vector diagram used to take hours in Illustrator. Today, a precise instruction, written or dictated, gets me animated, lightweight and sharp visuals, built directly into the code. The ones in this article were created that way.

Let me be clear, I do not hand my website over to AI. It publishes nothing without my approval, and I control every piece of data that goes online. Figures, dates and references are checked at the source before appearing on a page, and every change is recorded in the Git history, which lets me know who changed what and roll back at any time. The AI executes, I decide.

What it gives me is time, an enormous amount of time. Tasks that used to take me hours, such as laying out a page, producing a diagram or adapting content, now take a few minutes. I devote that time to what really gives my work its value: thinking, creativity, field experience and supporting my clients.

From SEO to GEO with the G.A.I.A method

For a very long time, I achieved my best SEO results with WordPress. But the web is entering a new era, that of GEO (Generative Engine Optimization), where generative engines and conversational agents are redefining how people search for information. It was to respond to this shift that, back in 2025, I formalised the G.A.I.A method, for Generative, Authority, Information and Agents.

The A for Agents is particularly fundamental. Tomorrow, searches will no longer be made mostly by humans clicking from link to link. A user will ask their personal agent to find them a renovation company, book a train ticket or buy a handbag. Robots will go and consult websites created and fed by other robots.

Google is starting to measure this. Since May 2026, Lighthouse, the Chrome audit tool behind PageSpeed Insights, has offered an agentic browsing category. It is still experimental and relies on standards that are still being drafted, but it already checks that an AI agent can read a page and use it. 7days.dev passes both checks that apply to it. In this ecosystem where AI talks to AI, a website must be clean, ultra-fast, semantically flawless and immediately readable by machines. The technical heaviness of a traditional CMS becomes a competitive disadvantage.

WordPress or Astro, the technical and financial balance sheet

Getting excellent performance from WordPress is possible, I have done it on many sites, but sometimes at the cost of considerable effort and constant maintenance. To optimise a WordPress site, you generally need to

• install and configure a caching plugin, some of which are paid, such as WP Rocket;
• pay for an image optimisation service, such as Imagify, or convert all your media to WebP or SVG yourself;
• defer the loading of heavy scripts, such as Google Tag Manager, Google Analytics or ad pixels;
• minify and clean up the CSS and JavaScript, self-host the fonts, and so on.

All these steps take time and technical skills, and often generate recurring subscriptions for plugins. With Astro, most of this optimisation is built in. The framework produces pure HTML and CSS, ready to be served.

On the hosting side, the cost becomes almost negligible. I use a virtual private server (VPS) I already had, since it also runs my automations and backs up my infrastructure, so 7days.dev costs me nothing extra. For someone starting from scratch, a VPS of this kind costs less than 10 euros a month with a host such as Hetzner, and services such as Cloudflare Pages, Netlify or Vercel even host a modest static site for free. Compared with a WordPress site, its premium plugins and its maintenance, the difference is considerable.

Can WordPress make the shift to AI?

The site you are reading runs on WordPress, and yet I hardly manage it from its back office any more. Thanks to the MCP protocol (Model Context Protocol), which lets an AI connect to a piece of software and act in it, I give my instructions to Claude, and it creates the pages, updates the texts, prepares the images and fills in their fields directly in WordPress. This article was put online that way.

In other words, on WordPress as on Astro, I now work in the same way. I talk to the AI, it executes, I approve. The back office becomes a fallback interface. What loses its purpose, then, is the CMS as a dashboard, not necessarily WordPress.

WordPress has clearly understood this. Since version 6.9, its core has included an Abilities API, which describes what the site can do so that AI agents can discover and use it, and its dedicated AI team is developing an official MCP adapter. Tomorrow, WordPress could become a content engine driven by agents rather than a dashboard. The challenge is considerable, adapting a twenty-year-old foundation to the age of agents, but if it rises to it, it will stay in the race.

Should you ditch WordPress, then?

Absolutely not, and this is not about crude disparagement. Many of my clients run on WordPress today, and the aim is not to tear everything down or launch systematic rebuilds. A website that works well, generates revenue and is properly maintained has no reason to be rebuilt on principle.

If oli-via-net.fr stays on WordPress, then, it is a deliberate choice. It works, it generates enquiries, and I do not carry out a rebuild for the sake of theory. The back office is also reassuring for many clients who are not yet familiar with managing a website through artificial intelligence, and WordPress retains real strengths for e-commerce or specific features.

On the other hand, when I design a new project from a blank page, offering WordPress now seems to me a default choice rather than a vision of the future. If a client is ready to take the step to a modern architecture, a foundation built on Astro and AI offers far greater performance and longevity for the era of the web that is opening up before us.

What will a website look like in 2030?

I do not claim to read the future, but the direction is already taking shape. A business owner will tell their AI assistant to find a tradesperson available on Thursday, compare three quotes and book the best one. That assistant will query companies’ websites directly, and sometimes their own agent, which will answer on availability, prices or lead times.

On the web of tomorrow, a website will have two audiences. Humans, who will come looking for trust, proof and personality. And agents, who will look for structured, up-to-date information and possible actions, booking, requesting a quote, checking stock. The standards Google is experimenting with, such as WebMCP or the llms.txt file, are designed precisely to prepare for this second audience.

The question will therefore no longer be which CMS to use, but whether an agent understands your website and can use it. A lightweight, fast site with clean code and verifiable content starts with a head start. That is exactly what I am building today, for my own sites as for my clients’.

Wondering whether your website should evolve, or whether your next project deserves a modern architecture? Fill in the form below to get a free audit.

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

Are these Trustpilot reviews real? Many people ask themselves that question. Among my clients, I often meet business owners who find reviews on their listing that they cannot trace, very negative ones, or on the contrary, surprisingly glowing ones on a competitor’s listing, and who are convinced they are fake.

The problem is that a conviction is not enough. Reporting a review to Trustpilot without a solid reason almost always ends in a refusal. To get a review removed, you need facts: verifiable evidence that the platform can recalculate itself, and that leaves no room for interpretation.

There are several ways to build that kind of evidence. I am going to share one of the methods I use, which I consider essential: timestamps. It does not require any secret tool, just some method, logic and rigour. To illustrate it, I will use a real case: a restaurant hit by a series of one-star reviews.

Trustpilot timestamps everything, to the second

Every account, every review and every business listing on Trustpilot has a 24-character ID. Its first 8 characters encode the exact moment it was created, in seconds. A conversion from hexadecimal to decimal is all it takes, a spreadsheet will do. It is raw data: it does not depend on what the author displays, and anyone can recalculate it from the public pages.

By timestamping every account, review and listing involved, I was able to rebuild the full timeline. Three clues came out of it.

Clue #1: two listings created 58 seconds apart

Something many businesses do not know: anyone can create a business’s listing on Trustpilot, without its consent, simply to post a review on it. The listing then exists online without the business knowing. If it finds out, it can ask to take over the listing: this is called claiming it. Until it does, the listing stays “unclaimed”, and the business can neither reply to reviews nor report them from its own account.

That was the case for this restaurant: it had never created its listing, and only discovered it four months later. Its ID gives the exact time it was created. And 58 seconds later, another listing was created: that of a competing restaurant, which was about to receive a flood of five-star reviews.

Another telling detail: both listings were created from Eastern Europe, while every other restaurant in the area was correctly located in France. So this was not a default setting, but the fingerprint of whoever created both listings. That same morning, the first five five-star reviews landed on the competitor’s listing. Three days later, the first one-star reviews arrived on the restaurant’s.

Clue #2: 48 throwaway accounts, made in clusters

By widening the analysis to neighbouring restaurants, I identified 48 accounts used for a single review, then abandoned, spread across four listings: 27 to inflate the rating of the competitor that benefited, 9 and 6 to attack two other restaurants, and 6 against my client. Almost all of them posted their review less than 30 seconds after the account was created. No customer writes a detailed review in a few seconds: the text was prepared in advance, then pasted. An account created just before its review can happen with a genuine customer; it is the pattern repeating, account after account, that gives the operation away.

Above all, these accounts are not created at random. They are made in clusters: on a single day, eight accounts were created in under two hours for the listing that benefited. And a writing quirk kept coming back from one review to the next, such as the targeted restaurant’s name typed entirely in capital letters.

The logic of the operation then becomes clear: one restaurant showing nothing but positive reviews, and around it, as if by chance, several competitors receiving negative ones. That is also what makes the method useful both ways: it helps defend a listing under attack, but also spot a business whose positive reviews are suspicious, including a competitor’s.

Clue #3: restaurants hit within the same window

This is the clue that leaves no room for doubt. One day, three competing restaurants were hit with one-star reviews, from accounts all created within 12 minutes. About ten days later, within 18 minutes, two accounts praised the restaurant that benefited and a third attacked another one.

A genuinely unhappy customer does not create their account within the same few minutes as accounts that, elsewhere, praise one competitor and attack others. This synchronisation across businesses is the signature of an operation run by a single actor, not of independent customer feedback. To make sure these coincidences are not down to chance, I run them through a permutation test: I shuffle the dates at random 2,000 times, and such a concentration almost never reappears.

The result: 42 fake reviews out of 42 removed

I documented everything in a file sent to Trustpilot: the full timeline, the IDs and the calculations, which the platform can redo itself. A few weeks later, on the three competing listings, all 42 fake reviews identified had been removed: the 27 five-star reviews of the restaurant that benefited, and the 15 one-star reviews of the two restaurants under attack. On the listing that benefited, the only review left was precisely the one the file identified as genuine.

For my client’s listing, one of the platform’s rules came into play: its reviews had been reported as soon as they were posted, before I got involved, but under the wrong reasons. Those reports had been rejected, and after one year, a review can no longer be reported, whatever the reason. The remaining route is a written exchange with the platform, which takes longer.

My method

1. Act fast. The one-year limit is a hard one. Ideally, report within days of the review being posted.
2. Timestamp everything you can. Accounts, reviews and listings: the IDs give the exact timeline, which the platform can recalculate.
3. Widen the search to neighbouring listings. A fake review operation rarely targets a single business. The evidence often shows up on competitors’ listings.
4. Build a body of evidence, not a single clue. Account clusters, synchronisation across targets, shared fingerprints, a permutation test, and always a baseline for the sector.
5. Pick the exact reason and follow the procedure. The wrong reason gets the report rejected. And the request must come from the business’s own domain email address, with no one else in copy.
6. Never report a genuine customer. A genuinely unhappy customer deserves a careful public reply. That is also what gives credibility to the reports that are well founded.

Timestamps are not read review by review: they are read by comparing accounts and listings with one another. That is where an organised operation, however discreet, always ends up leaving a trace.

Do some Trustpilot or Google reviews look suspicious to you? I analyse them and tell you frankly what can be demonstrated, what cannot, and how to respond.

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

It is a situation I come across regularly with the SME owners who contact me: a WordPress site goes down, and there is only one question on their mind, “how much to fix it?”. Here is how I handled the case on my own site, what I chose, what I ruled out, and what the measurements show.

The finding: an outage that cannot be repaired

On 2 September, my check left no room for doubt: HTTP 500 on 7days.dev, still a 500 two days later. The site shared a shared-hosting plan with a dozen other sites, some of which took more than 20 seconds to respond that day.

A WordPress site is two things: files, and a database that holds all the content. Here, the files were there but the database was not. “Repairing” therefore meant reinstalling an empty WordPress and re-entering everything. At that cost, the real question becomes: is WordPress still the right tool for this site?

The decision: what I chose, and what I ruled out

7days.dev is our division dedicated to business process automation. It is made up of presentation pages and articles. No basket, no customer area, no complex forms. Before deciding, I compared the options on that specific criterion.

• Rebuilding a WordPress site: ruled out. The same database to host and back up, the same plugin updates, the same risk tomorrow.
• A third-party hosted “headless” CMS (such as Sanity): ruled out. The content would live with an outside provider, and I want to stay in control of my data.
• Next.js: ruled out. Excellent for an application, oversized for a site with no dynamic needs.
• Astro, as a static site: chosen. It is the go-to tool for content sites in 2026. It produces ready-made HTML pages, with no database and no code executed on each visit.

I accepted the trade-off from the outset: no more WordPress admin interface. The content is edited as text files. For me that is not a problem, since my automation and AI tools write directly to those files. For a client who wants to edit their own pages every week, it is a real factor to weigh up.

What I put in place

1. Content in Markdown in a Git repository. Each page is a versioned text file. The full history is kept, and the site can be rebuilt identically in a few minutes.
2. An automatic build on every push. When I push a change, the site is recompiled in an isolated container, then the new version replaces the old one in one go, with no visible intermediate state.
3. A server that only serves files. No code is executed on a visit, so there is simply nothing to hack on the server side. HTTPS certificates renew themselves.
4. An automatic publishing chain. An article dropped in by one of my workflows goes live with no intervention. Measured on 16 September: pushed at 11:02:59, live at 11:03:06, in other words 7 seconds.
5. Animated vector visuals. Instead of stock photos, 16 animated SVG diagrams drawn in code, lightweight and sharp on every screen.

The content, however, is anything but automatic. I reread every page on screen during the rebuild, and every fact stated (date, partner, technical choice) must be backed by a source, an official register or a press article. If I do not have the source, the sentence does not go online.

The measured results

Before the public launch, I had a bot go through all 105 pages of the site: zero JavaScript errors, zero overflow on a 390-pixel phone screen, and a 10 KB compressed stylesheet.

On the SEO side, I validated the structured data in Google’s official tool before expecting any traffic at all. Breadcrumbs, organisation and services are all recognised without errors.

First signals in Search Console

The first impressions appeared on 20 September. Eight days later, Search Console counted 521 impressions and 4 clicks, at an average position of 14.5. Above all, the curve is rising fast: 11 impressions on 20 September, 153 on the 26th.

What interests me most are the queries. Apart from the brand “7days” (position 6), they are very specific searches, made by business owners who already have a problem in mind: “automatisation des relances clients” (automating customer reminders), “automatisation des notes de frais” (automating expense reports), “saisie automatique des factures” (automatic invoice entry), “devis automatisé” (automated quotes). The pages that capture the most impressions are those about chasing unpaid invoices (42), entering supplier invoices (34) and recurring invoicing (21). That is exactly the site’s target audience.

Eight days is a start, not a trend. I will review it again at one month and then at three months, with the same indicators.

Do WordPress and the back office still make sense in the age of AI?

That is the real question this project raised for me. For years, the back office was the only way for a non-developer to edit their website. Today, a well-equipped AI assistant can write a page, fix a typo or publish an article from a single sentence, directly in the files or through the site’s API. For a content site whose changes go through me and my tools, the admin interface becomes optional.

That does not make WordPress an outdated solution. The article you are reading is published on oli-via-net.fr, which runs on WordPress, and it was placed there by those same tools, through the API. The two approaches coexist very well. In my view, WordPress and its back office remain the right choice whenever you need:

• to sell online: catalogue, stock, payments and orders with WooCommerce;
• specific features: booking, members’ area, advanced forms, industry-specific plugins;
• autonomy on the client side: several people publishing, with roles and approval;
• an ecosystem: thousands of plugins and providers able to take over the site.

A static site, for its part, really comes into its own for a showcase or editorial site whose content changes a few times a month, where speed and security come first and publishing can be automated. And in every case, a database without a tested backup is an outage waiting to happen. It is the first thing I check during an audit.

Is your website down or slow, or are you unsure whether to develop it or rebuild it? I will look at your situation and tell you frankly what makes sense, WordPress included.

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

A major change is rewriting the rules.

Starting a business in 2026 is therefore no longer just about choosing a name and launching a website. You need to think from the outset about how it will fit with the Google ecosystem. In this article, I explain why, and above all which strategy to adopt.

It is no longer a matter of creating a business and then adapting its Google profile.
You have to do the reverse: think from the outset about a business name that is consistent with your future visibility on Google, then structure your business around that name, making sure you remain credible, consistent and in line with Google’s expectations.

Google has strengthened and automated its control systems, and this is profoundly changing the way businesses must manage their local visibility.

Today, in many sectors, the Google profile generates a significant share of calls, quote requests and in-store visits. It often performs better than a website.

Your business name: an underestimated strategic mistake (and often an irreversible one)

This is the most critical point… and the most often overlooked.

Many businesses try to optimise their Google profile by adding keywords, a city or a description of their activity directly to the name. For years, this approach helped improve local rankings.

The result: more and more Google Business Profiles are being suspended overnight, sometimes with the total loss of the reviews they had built up. For some businesses, this means an immediate collapse in customer acquisition.

That way of working therefore belongs to the past. With automated checks, these practices are now detected and considered non-compliant. Google expects a clear match between the name displayed, the administrative reality and the presence on the ground.

This means your name must be consistent with:

• your registered company name
• your shop sign (if you have one)
• your real presence (physically visible or verifiable)

A real case: Google changes the name automatically

To illustrate this concretely, here is a real case I handled recently.

A bathroom specialist in Paris, trading under the name “Les Bains d’Alexandre”, had a properly optimised Google profile for several years.

Following a report by a competitor or an automatic update, Google changed the profile’s name and replaced it with the company’s legal name, “Décoration Générale”.

👉 This kind of change is becoming more and more frequent.

Google can cross-reference data from various official databases, including:

• INSEE
• Infogreffe
• Papers
• Société.com

When the name displayed on the profile does not clearly match the administrative information, the system may decide to correct the title automatically.

Why the name was changed

In this specific case, Google considered that:

• the name “Les Bains d’Alexandre” was not sufficiently linked to the official data
• the name “Décoration Générale” was the main reference

The result: the profile was changed without any action by its owner.

How we restored the name

We were able to restore the name “Les Bains d’Alexandre” for two essential reasons.

First, it really is the establishment’s declared shop sign name, which makes it legitimate for display on the profile.

Second, we provided clear visual proof:
a photo of the shopfront showing the “Les Bains d’Alexandre” sign, visible from the street.

👉 This point is decisive.

Google does not rely only on administrative data, but also on the reality on the ground. A visible, verifiable shop sign becomes a strong signal of legitimacy.

What to take away from this case

This case perfectly illustrates the current limits of the system.

If the business had not had:

• a real shop sign
• consistency between its trading name and its activity
• usable visual proof

it would have been very hard to get back to the original name.

In some cases, the only solution is even to change the shop sign’s name administratively so that all the signals line up:

• legal data
• physical presence
• Google profile

The real strategy therefore consists of planning ahead from the moment the business is created.

Let us take a simple example:

A tradesperson creates a company called “Martin Services” and then tries to appear on Google as “Plombier Paris Urgence” (emergency plumber Paris). This inconsistency becomes a negative signal.

Conversely, a business owner who chooses a shop sign such as “Plombier Paris Martin” from the start aligns:

• their commercial identity
• their Google profile
• their local visibility

The real strategy I recommend, and which seems to me increasingly unavoidable, is to think from the outset about the name you will actually use in your business (shop sign, communication, website…) and to make sure it will be consistent with your future visibility on Google. The aim is to align your registered company name, your shop sign, your presence on the ground and your local visibility from the start. Ideally, the sign displayed on site should be consistent with the name used on your Google profile, without trying to over-optimise or create an artificial gap.

Google no longer takes you at your word:

As a senior SEO consultant, I have been observing a very clear change in Google’s control systems for several years. We have now entered a logic in which data is cross-referenced on a large scale and a business’s overall consistency is systematically analysed. Algorithms now cross-reference a considerable volume of data:

• official registers (INSEE, Infogreffe, public databases)
• the content of your website
• the information on your Google profile
• your legal notice
• user signals
• visual elements (photos, Street View)

The aim is simple: to check that your business really exists as it is presented.

This implies a fundamental change:
it is no longer possible to build local visibility on an approximation or an “optimised” strategy.

For example, claiming to be based in Paris while operating from Orléans becomes very hard to sustain. Inconsistencies are detected quickly, notably through data correlation.

Google no longer just seeks to index, it seeks to validate the overall credibility of an entity.

Customer reviews: the end of artificial strategies

The same phenomenon applies to reviews. How to avoid negative reviews, or report fake reviews.

For a long time, some businesses tried to influence their reputation by asking friends and family or by generating indirect reviews.

Google can now cross-reference user accounts, their interaction histories, their browsing behaviour and their location data, which lets it identify signals of consistency or, conversely, anomalies.

A review left by a user who has never been in a geographical area consistent with your activity becomes a weak, even suspicious, signal. Likewise, repetitive patterns or close links between accounts can be detected.

What Google is trying to measure is no longer just a volume of reviews, but the consistency between the experience the user describes and the reality observed through their signals.

Physical address: an increasingly checked criterion

Showing an address on Google Maps is a powerful lever, but it is regulated.

To be eligible, a business must be able to welcome customers in real conditions:

• an accessible place
• an identifiable activity
• visible signage

With tools such as Street View, photos or verification videos, Google can check these elements.

Fictitious addresses, registered-office addresses with no real activity and simulated presences are increasingly detected. In time, these practices are set to disappear completely.

Service area: a strategic alternative

For service providers, showing an address is not always relevant. In many cases, it is better to use a service area.

This model lets you define an area of operation consistent with the business’s real activity. It avoids the inconsistencies linked to a non-compliant address while keeping good local visibility.

Google now recommends this approach in many cases, especially for businesses that do not welcome customers on site.

Automated checks: a paradigm shift

The most important change remains automation.

Google now uses advanced analysis systems combining artificial intelligence, image recognition and behavioural analysis. These systems can identify inconsistencies, detect anomalies and automatically trigger checks when suspicious signals are detected.

What is changing is not only the level of scrutiny, but its speed.

Where some practices used to go unnoticed for months, they can now be detected within a few days.

Profile changes: an underestimated risk factor

Another critical point concerns changes.

Regularly changing a profile’s name, category or key information can send negative signals to Google’s control systems.

When a profile starts to be monitored, each additional change raises the level of scrutiny. That is often when suspensions happen.

An effective strategy therefore rests on:

• a solid initial set-up
• consistency over time
• a measured approach to changes

Conclusion

Google Business Profile is no longer just a visibility lever. It is now a system for validating a business’s credibility. Name, address, activity, reviews: everything is analysed, cross-referenced and checked.

Artificial optimisation strategies are gradually disappearing in favour of a model based on consistency and transparency. In this context, businesses that plan ahead from the moment they are created gain a considerable lead. The others expose themselves to blocks, suspensions and a sudden loss of visibility.

The difference is no longer made on technique alone, it is also made on structure.

Sources & references

• Google Business Profile Guidelines (Google, updated 2025–2026)
• Google Search Central – Local SEO & Entity Validation (2025)
• BrightLocal – Local Consumer Review Survey 2025
• Whitespark – Local Ranking Factors 2025
• Statista – Local Search Behavior Europe 2025
• Deloitte – Digital Consumer Trends 2025
• McKinsey – The Future of Local Commerce 2025

Written by

Nicolas Peter, Web Performance, certified Google Ads expert in Paris and Perpignan
Oli-via-net agency

In 2026, French retail will probably go through a significant rebalancing. E-commerce continues its impressive growth, with €175.3 billion in 2024 according to FEVAD and a stated target of €200 billion in 2026, but it is no longer the undisputed dominant channel.

At the same time, physical shops are regaining ground, driven by a crisis of trust in online shopping and a technology ecosystem dominated by Google that is accelerating drive-to-store*. The physical shop no longer stands against digital, the two become complementary. And that is excellent news for local shops!

*Drive-to-store refers to digital strategies that aim to turn an online ad into a physical visit to a shop, building on local intent, proximity and actions such as directions or calls.

Fraud, stress, uncertainty: what is at stake for physical shops in 2026?

According to recent surveys (Havas Commerce/CSA 2024–2025, stable in 2025), 84% of French people still prefer to buy in a physical shop for many products, precisely for that peace of mind: seeing, touching and trying on the spot, getting immediate human advice, and solving a problem without a distant intermediary.

Players such as Shein have been at the centre of major controversies in Europe, combining non-compliant products, suspected circumvention of regulations and environmental impacts, to the point of fuelling draft laws aimed directly at ultra fast fashion. For its part, Vinted faces a growing number of reports of scams, fake profiles, off-platform payments and products presented in a misleading way, sometimes using AI-generated images.

These excesses, widely reported on social media and in the press, end up weighing heavily on the online reputation of the platforms concerned. For many consumers, they reinforce the idea that buying online now carries a higher perceived risk than buying in a shop, where the product is visible, the person you deal with is identifiable and the transaction is immediate.

According to Oneytrust (in partnership with FEVAD), online fraud now accounts for more than a billion euros in losses for French online retail every year, with around 60 million fraudulent transactions estimated. These figures mainly concern online retailers, but they feed straight through to buyers via sophisticated scams: smishing (fraudulent texts imitating couriers), fake shopping sites, counterfeits, AI-generated product photos designed to mislead on real quality, identity theft or payment fraud. At the same time, overall cybercrime hits French businesses to the tune of more than €100 billion in 2024 (Statista), with phishing as the main vector (60% of attacks according to the CESIN Barometer 2025), and 82% of e-commerce executives expecting an increase in attempts in 2025.

Beyond the direct financial losses, these incidents carry a psychological and time cost for the consumer: post-purchase stress (fear that the parcel will not arrive or will be counterfeit), time lost on disputes, complicated follow-ups, calls to customer services, and often lengthy refund procedures. Add to that returns that increasingly come at a cost.

The result: once the amount at stake exceeds a few hundred euros, the trade-off changes radically. The convenience of a click no longer outweighs the perceived risk, the uncertain wait (lead times, real quality, delivery) or the potential disappointment. For a high-tech product (smartphone, computer, tablet, camera, etc.), French consumers even accept a premium of 5% to 15% when buying in store compared with the online price, provided the difference is justified by concrete services or benefits.

The shop thus becomes a true sanctuary of trust again, where the sensory and human experience removes much of the uncertainty of online shopping. This is not a rejection of e-commerce, which remains dominant for recurring, low-risk purchases (consumer goods, restocking), but a rebalancing towards more security and peace of mind when the stakes are high. This search for human, immediate reliability largely explains why phygital and drive-to-store are gaining ground in 2025–2026.

What is “phygital”? A concept from 2013 that has not aged a day.

The customer searches online, compares on their phone, reads Google reviews, then goes to the shop to complete the purchase, where they can use click & collect or a payment made online. Physical does not push out online, it complements it and makes it more effective.

• In fashion, leather goods or accessories, trying things on remains essential and ROPO (Research Online, Purchase Offline) has become the norm.
• Crafts, one-off pieces and works of art draw their value from authenticity and from meeting the maker, so the shop becomes a genuine storytelling medium.
• For premium, technical or high-commitment products (furniture, hi-fi, jewellery, specialist sports gear), reassurance matters more than price alone.

The physical shop is no longer an isolated destination but the anchor point of a digital ecosystem that serves it and enhances it. It is becoming an experiential showroom, a service centre and an ultra-fast collection point. These arrangements turn the shop into the final point of conversion, even when the journey starts on Google, Maps or social media.

Which paid strategy pays off for local shops in 2026? A closer look at drive-to-store Google Ads

Unlike purely e-commerce approaches, drive-to-store campaigns do not try to maximise clicks but to generate qualified visits to the shop. In this logic, Google Ads becomes a key strategic lever for turning digital visibility into real footfall.

It relies on a precise combination: Performance Max campaigns linked to the Google Business Profile, local conversion goals (directions, calls, Maps interactions) and consistent geographical signals. This approach makes it possible to measure what was historically invisible, the direct impact of digital on physical traffic. It does, however, require a rigorous set-up of local conversion tracking and a different reading of performance, focused on footfall and not only on online revenue.

Google’s role in this rebalancing is therefore decisive. Its algorithm explicitly aims to turn high-intent local searches, such as “best sofa in Bordeaux”, into directions to relevant physical shops.

Genuine reviews play a key role here, acting as the structuring trusted third party of this new local economy. Far from weakening local retail, Google now offers it a digital arena in which to compete on a more level footing with the e-commerce giants, provided it masters the strategic levers.

Reviews, the new digital gold: why has Google made them its currency?

Thanks to the combined ecosystem of Google Maps, Google Business Profile and Performance Max, enriched since 2025 by the automatic integration of Waze Ads into drive-to-store strategies, the goal is clear: turn a search intent into a real journey to a point of sale.

This is precisely where Google plays a central role as a catalyst. Now well beyond its original function as a search engine, Google has established itself as an engine of local intent, a recommendation system, a trusted third party through reviews, and a genuine driver of physical traffic.

In this context, reviews become the new currency of local retail. The stakes are strategic for Google. Reliable reviews generate more trust, therefore more use, and ultimately more monetisation through directions, local campaigns and Maps interactions.

Recent algorithm changes reinforce this dynamic. In 2025, Google considerably improved its detection of fake reviews, relying on semantic analysis, behavioural consistency and the correlation between reviews, physical visits and directions. We all use Google Maps or Waze, so it is not hard for them to know whether someone really went to a given gym, or whether a given customer came to your offices. This approach mechanically favours authentic businesses that people actually visit and that are firmly rooted in their area.

For small shops, this change represents a historic opportunity: access to visibility and credibility comparable to those of the e-commerce giants, without disproportionate marketing budgets.

How AI narrows the gap between independents and large chains?

Artificial intelligence now acts as a major accelerator of this rebalancing, making available levers once reserved for large organisations:

• Predictive stock management
• Customer service automation
• Replies to reviews
• Data-driven local campaigns
• Production of optimised content…

An independent retailer can now reach a level of operational efficiency comparable to that of businesses with dedicated teams. The barrier to entry is gradually fading. Even Amazon is gradually losing its exclusive hold on large-scale optimisation and automation.

The real differentiator no longer lies in size or investment capacity, but in the ability to offer a human, personalised and consistent experience, now augmented by technological power within everyone’s reach.

Generative AI will amplify this movement. Tomorrow, conversational assistants and recommendation engines will no longer simply highlight the cheapest offer, but will favour the best-rated, nearby businesses offering a reliable customer experience. In this environment, a physical shop with good visibility naturally becomes “AI-compatible”, because it ticks the boxes of trust, proximity and social proof these systems look for.

Why can local shops be the big winners of AI?

The next stage of this evolution may well be played out with generative and conversational AI. These assistants, built into search, will no longer just list the cheapest links. They will tend to give concise, action-oriented answers.

Drawing heavily on data about trust, proximity and social proof (reviews), they could naturally favour a recommendation such as: “Shop X, 10 minutes from you, has the item in stock, is rated 4.8/5 with more than 200 reviews and offers a free fitting.”

This prospect confirms the decisive advantage of a well-rated physical shop with good visibility.

Conclusion

For retailers, the signal is unambiguous: the future lies neither in all-digital nor in all-physical, but in the ability to build a smooth, coherent ecosystem in which each channel strengthens the others.

In this new configuration, technology is no longer a disruptive force working against local retail. On the contrary, driven by Google and artificial intelligence, it becomes a lever for rebalancing and reassurance, capable of giving power back to local players in the face of global platforms.

Sources & references:

• FEVAD, French e-commerce figures 2024–2025
• FBI, Internet Crime Report 2024
• European Retail Outlook 2025
• Google Ads & Google Business Profile Updates 2025
• McKinsey, The Future of Omnichannel Retail (2025)
• Deloitte, Consumer Trust & Commerce Report (2025)
• Statista, In-store vs Online Retail Share Europe (2025)

Written by

Nicolas Peter, Web Performance, certified Google Ads expert in Paris and Perpignan
Oli-via-net agency

Your Google Business Profile (GBP) is the central stage where trust in your brand is won or lost. Before your website, before your shop, this is where 92% of consumers form their first impression (according to the Digital Trust report by BrightLocal).

We are going to look at a practical, pragmatic and sometimes counter-intuitive approach, in which a skilful settlement is often worth more than a war of attrition. I will also share two methods I have put in place to moderate your reviews effectively in 2026:

• the R.E.P.A.I.R method , which organises your whole strategy for handling a negative review (responsiveness, private exchange, resolution, analysis and continuous improvement)
• the 5 Rs rule, to write professional public replies, even when the review is unfair or emotional.

The R.E.P.A.I.R method and the 5 Rs rule: your anti-negative-review action plan for 2026

To help you through the review moderation process (where, when, how), here is the R.E.P.A.I.R method, workable in any situation.

R – Receive & Recognise (instantly): never leave a negative review without an acknowledgement. A short public reply within 2 hours shows you are paying attention. For that, lean on the 5 Rs rule, explained below.

E – Explore in private (within 24h): open a dialogue away from the public square. Listen.

P – Propose a proportionate solution (within 72h): this is where the “amicable settlement” happens — your best return on investment in 2026.

A – Analyse with AI (ChatGPT, Gemini, Claude…): use the free tools to audit the overall sentiment and refine your standard replies.

I – Integrate the “lesson” (internally): every negative review you resolve should lead to a process improvement (a new check, some training, a tool, a new recipe…).

R – Report back publicly (where possible): if the conflict is resolved and the customer agrees, add a note to your public reply: “Following our conversation, we are glad to have been able to resolve this with you. Thank you again for your feedback.” It is the ultimate proof of how well you handle things.

How to write your reply to a negative review: the 5 Rs rule

Before talking about amicable settlements, artificial intelligence or overall strategy, one essential point needs clearing up: how to reply well to a negative review.

In 2026, a reply to a Google review is not a simple courtesy message. It is a public act, read by dozens — sometimes hundreds — of future customers.
That is why I use a simple, teachable, repeatable structure I call the 5 Rs method.

It lets you structure an intelligent message, readable and reassuring:

1. Reflect the emotion
Show that you have understood how they felt: “We understand your frustration at having to wait…”

2. Recognise the fact (when it is verifiable)
Confirm the objective element: “You are right, that wait was abnormally long that day.”

3. Take Responsibility
Own it without making excuses: “It is our responsibility to guarantee a smooth experience.”

4. Resolve
Offer a concrete solution (often away from the public thread): “Please get in touch so that we can find a suitable solution.”

5. Reform
Show that the mistake is being used to improve your processes: “This incident has led us to review our welcome procedure.”

The amicable settlement – your best ROI (return on investment) in 2026

The first instinct should be to treat a negative review not as a stain to be erased but as an opportunity for direct contact. The race to file technical reports — often long, costly, and decided by the platforms — is being overtaken by a logic of relational ROI.

Why the amicable route has become unavoidable

A ReviewTrackers study (Q4 2025) showed that 68% of 1- or 2-star negative reviews come from a misunderstanding or a one-off incident, not from structural incompetence. More striking still: 51% of customers who posted such a review agree to change or delete it if the business offers a personal, sincere resolution within 72 hours.

The arithmetic is simple: compare the cost of a targeted goodwill gesture with the hidden cost of a negative review that stays online — potential customers lost, a lower conversion rate on your profile, a degraded trust score in Google’s local algorithm.

Concrete examples & profitable scenarios

Scenario A – the restaurateur and the unhappy customer:

• The review: “Disappointed! Booked for a birthday, table given to us 25 minutes late, the whole mood was ruined.”
• The “defensive” approach (removal): time spent documenting, reporting, following up. Agency cost: €300 minimum. Uncertain outcome, customer lost for good.
• The “R.E.P.A.I.R” approach (amicable): the manager identifies the booking (thanks to the detail) and calls the customer the next day, apologises personally, acknowledges the failure at the door, and offers a complimentary bottle of champagne for their next visit. Cost: €35.
• Result: the customer, touched by the gesture, changes the review to 4 stars: “After my review the manager called me personally. Human error, exemplary response. Giving them another chance!” The overall rating is preserved, and the public reply shows everyone how well you handle after-sales care.

Scenario B – the tradesperson and the detail that grates:

• The review: “Work generally fine but sloppy finishing on the skirting board. No answer to my follow-up.”
• The “defensive” approach: a public reply (“The finishing matched the quote”) that antagonises the customer and puts future prospects on alert.
• The “5 Rs rule” approach: private message: “Mr Dupont, we have received your feedback and thank you for it. You are right, that detail is not up to our standard. One of our team will come by tomorrow between 8 and 9 to put it right, at no charge. Our apologies for the slow reply.” Cost: one hour of labour.
• Result: the customer is pleasantly surprised. They often update the review to highlight the responsiveness and professionalism, turning a negative point into a demonstration of reliability.

The timing of an amicable settlement (the 24/72 rule):

Within 24 hours: acknowledge it publicly (“We take your feedback very seriously and are contacting you privately.”). That stops the emotional escalation.

Within 72 hours: close the amicable settlement. Beyond that, the customer concludes that you are not prioritising their problem.

The “reflex” replies that make everything worse… and the ones that turn it around

A real case: a car garage – 1-star review.
“Botched repair! The car still makes the same noise. They have no idea what they are doing.”

🛑 The self-centred reaction (badly received):

“We dispute this review. Our mechanics are qualified and the work was carried out in line with the diagnosis. The problem must lie elsewhere.”

The effect: you come across as arrogant, you confirm the assumed incompetence, and you put off every future customer.

✅ A reply using the 5 Rs rule:

“Hello Mr Martin, we are sincerely sorry to hear that the noise persists despite our work. We understand your disappointment (Reflect). You are right, the problem should have been resolved by the repair (Recognise). It is our responsibility to guarantee the quality of our work (Responsibility). Could you call us back on 01.XX.XX.XX.XX so that we can schedule a full diagnosis and an immediate re-repair at our expense? (Resolve). Feedback like yours helps us strengthen our final quality checks (Reform).”

The effect: you come across as professional, straight and customer-focused. Readers think: “Even when things go wrong, they handle it well.”

Replying to reviews: how AI can help you… and how to avoid the traps

In 2026 generative AI is everywhere, and anyone can use it. The trap is using it as a simple copy-paste for replies that ring false. The right approach is to make it a co-pilot: a tool that helps you analyse, phrase and reply more intelligently, while keeping your human tone.

Since a 2025 study estimated that around 40% of online reviews show signs of AI involvement, moderating your Google Business Profile is no longer just about “replying politely”. It has become a real strategic game: understanding the intent, staying sharp on the legal framework, and replying intelligently enough to reassure your future customers (the ones reading in silence).

Here is a very concrete example of the new reality of Google reviews in 2026: the customer used an AI to write the review… and accidentally left in the tell-tale sentence “Would you like me to adapt it a little more, or make it more formal?”
That detail shows how far reviews can now be influenced (or generated) artificially — and above all, how completely a clumsy reply can miss the point.

“AI should help you structure, but the final validation has to be human to stay compliant with Google’s guidelines on AI-generated content”. It is also what keeps you compliant with Google’s latest AI policies (AI Content Policy).

Here, the owner fixates on that “question” and replies beside the point, instead of applying the 5 Rs rule: acknowledge the disappointment, take responsibility, offer a solution and reassure future customers.
It is exactly the kind of situation where AI becomes absurd when it is used badly… whereas it can also be a formidable tool when it is used intelligently.

Here is how to use AI intelligently to reply to Google reviews

Google’s AI (Search Generative Experience) is evolving fast. In 2026 it is starting to spot replies generated 100% by AI that lack contextual specificity. The penalty? A downgrade in local visibility, because Google favours authentic interactions. AI serves here as a writing aid, but every reply has to be personalised with real contextual elements to guarantee complete transparency towards users. Always add a unique detail (the product name, the day of the week, a reference to your internal process).

For example: instead of “We are sorry”, write “We are sorry about the wait on Saturday morning; we had a staffing problem and we have since fixed the rota.”

It reads as genuine, human, and the opposite of generic AI.

The 3 AI habits to adopt when replying to Google reviews:

The sentiment radar: ➡️ Aim: understand what really annoys customers (without assuming)

Paste your last 50 reviews into a tool such as ChatGPT with this prompt: “Analyse this set of customer reviews for a [your sector]. List the 3 dominant emotions (frustration, disappointment, satisfaction, say) and identify the 2 most specific recurring friction points, quoting precise examples.” You may discover that “being kept waiting on the phone” is a far bigger problem than the quality of your product.

The anti-hot-reply filter: ➡️ Aim: avoid replying in the heat of the moment.

Before publishing a reply written in anger, put it to the AI: “As a customer reading this reply to a negative review, what main feeling does it convey to me (irritation, contempt, empathy, professionalism)? Suggest a calmer rewrite that keeps the factual elements [list them].”

The training scenario generator ➡️ Aim: rehearse your reply before the real crisis:

Ask: “Generate 5 realistic, difficult negative reviews for a [your trade], including an emotional customer, a technical customer and a customer who is in the wrong. For each case, provide a model reply using the 5 Rs rule (paste the section of this article explaining the rule into your prompt).” Use those scenarios to train your team.

Conclusion: what your future customers remember is not the review… it is your reply

In 2026, moderating your Google Business Profile is no longer a question of “reputation management” in the passive sense. It is an active discipline of growth and customer relations. It means understanding that the cost of a well-placed goodwill gesture is often the most profitable marketing there is. And it means accepting that AI is a formidable amplifier, provided you keep a human hand on the wheel of authenticity.

By adopting the R.E.P.A.I.R method and structuring your replies with the 5 Rs rule, you give tangible proof that your business is trustworthy, reliable and fundamentally human.

Going further: this article covered “reactive”, relationship-led moderation. The next instalment will unpack the technical 2026 review reporting process: when and how to make it succeed against Google’s ever stricter algorithms.

Sources & references:

• BrightLocal, “Digital Trust & Local Reviews Report”, 2026.
• E-Reputation Insights, “Barometer of Online Review Authenticity”, 2025.
• ReviewTrackers, “Customer Recovery & Review Dynamics”, Q4 2025.
• Local SEO Watch, “Impact of Manager Responses on Local CVR”, 2026.
• Local Trust Report, 2026.
• Official Journal of the European Union, “Directive (EU) 2026/… amending Directive 2019/2161”, 2026.
• EU Artificial Intelligence Regulation (AI Act), phased application 2024-2026.

Written by

Nicolas Peter – Web performance, Google-certified Ads & Google Business Profile strategy expert in Paris and Perpignan
Oli-via-net agency

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

As a digital marketing professional for more than 15 years, I have had a front-row seat on the breakneck evolution of artificial intelligence (AI). For two years Silicon Valley has been selling us an almost messianic vision: a single, omnipotent artificial intelligence, able to do everything better than a human.

But what if that vision were a decoy? What if, as in medicine, real excellence lay in specialisation?

My conviction, backed by technical and economic analysis, is that we are watching not the birth of a monopoly but an inevitable fragmentation of the AI landscape. That obviously calls into question the current overvaluation of the “Magnificent Seven” (Apple, Microsoft, Alphabet, Amazon, Meta, Tesla and Nvidia), who are positioning themselves as the future major players in a sector that remains speculative, fuelled by circular investment and facing colossal energy challenges. Let us look at it together.

Updated 4 September 2026. I published this article on 2 January 2026 with a market scenario that stopped in March. Eight months on, I am updating it for three reasons. The fragmentation I predicted happened far faster than I expected. The catalysts I had dated came and went without causing a break, which deserves to be said just as plainly as the rest. And above all, the main risk factor has changed in nature.

The myth of the “universal doctor”: why a “generalist” AI cannot do everything

An AI that would be surgeon, radiologist, psychologist and general practitioner all at once — the digital “universal doctor”. The financial markets, fascinated, pushed the valuations of AI’s “Seven Wonders” (Nvidia, Microsoft-OpenAI, Google and the rest) to stratospheric levels, anticipating a future monopoly.

Think of AI as the medical world. A general practitioner is valuable for an initial diagnosis and an overall view, but for complex problems you turn to specialists: a cardiologist for the heart, a neurologist for the brain, an oncologist for cancer. In the same way, a generalist AI such as ChatGPT (OpenAI) excels at everyday tasks — basic writing, brainstorming — but struggles with sharper needs, where specialised AIs outperform it.

Midjourney: still a premium reference for artistic rendering and for an “aesthetic signature” much appreciated by creatives, with a Discord ecosystem that remains active. But its leadership is no longer beyond question: the image models built into Gemini and ChatGPT have largely closed the gap in the independent 2026 rankings.

Claude (Anthropic): the specialist in programming and long-form writing. It stands out for its precision on complex code (debugging advanced algorithms, for instance) and its ability to structure in-depth content, with fewer hallucinations. It is the clearest riser of the year: from 2% market share at the end of 2025 to around 10% by the summer of 2026, that is close to 245 million monthly users, driven by professional use.

Perplexity : AI’s “academic researcher”, expert at reliable sourcing and citations. It is also the most useful counter-demonstration of 2026: its share fell from 5.3% in January to around 2% in August, simply because ChatGPT, Gemini and Claude built sourced search in natively. Specialisation protects you only as long as a generalist cannot absorb it. That is the nuance I did not state firmly enough in January, and it is worth correcting.

Grok (xAI): the specialist in instant information and social trend analysis through X (formerly Twitter). Perfect for insight on real-time events. Its share has held steady at around 2.8% in 2026, in a market that has otherwise moved a great deal — its niche protects it as much as it caps it.

Gemini (Google) and AI Mode: the big winner of 2026. A champion of multimodality, but above all of distribution: Search, Workspace, Android, Gmail. Its share went from 18.2% at the end of 2025 to 27.6% in September 2026, with the strongest quarterly growth in the sector (+17%). It is not winning by being better, it is winning by already being there — which, for anyone who works in search, is no surprise at all.

And why not a super generalist AI that would absorb the others?

Optimising for everything dilutes excellence. Just as a doctor cannot master every speciality without years of study, scaling an AI to excel everywhere demands massive resources in data and compute. Niches (secure code, sourced research) need dedicated architectures.

The figures confirm it, and faster than I was writing in January: ChatGPT went from 87% of the market in 2024 to 68-74% at the end of 2025, then to 51.5% in September 2026 according to First Page Sage. Another measure, Sensor Tower’s, centred on mobile usage, already put it below the 50% mark in May 2026, at 46.4%. The methodologies differ; the direction does not.

One point deserves to be made clear, because it is consistently misread: in absolute terms ChatGPT is still growing, with roughly 1.1 billion monthly users against 662 million for Gemini and 245 million for Claude. It is not shrinking, it is growing more slowly than the market. That is the exact signature of a loss of hegemony, and it is far harder to correct than a drop in visitors.

An alternative, and a new challenge: “the meta-engine that aggregates the AIs”

Rather than switching between 10 AIs, imagine a “new Google”: a search engine that queries several AIs in parallel, assesses their credibility, compares the answers to surface alternatives, and ranks the whole thing intelligently.

The results page would mix organic search (organic links) and AI blocks: a title, a short description, and a link to the full answer. That would solve fragmentation without a monopoly, and democratise AI without depending on a single giant. As things stand, though, my vision of an AI meta-engine runs into three major locks.

Technically, the cumulative latency and the synthesis of heterogeneous answers make the experience slow and confusing.

Economically, the cost per query, multiplied across several models, is prohibitive at scale, with no viable business model. Could the arrival of high-performing open-source models change that by making aggregation feasible?

Strategically, will the AI giants ever accept being content suppliers inside a meta-engine? And would a meta-engine not itself become a more powerful monopoly than the ones it claims to bypass? Only time will tell.

A speculative bubble, circular investment and overvaluation of the Seven Wonders?

The Magnificent Seven are selling themselves as the future giants of AI. Take an example you all know: Nvidia, the leader in the GPU chips that AI depends on. Its share price exploded on the back of circular deals with OpenAI. Nvidia sells billions in chips to OpenAI, which attracts massive investment, which in turn inflates demand for Nvidia chips.

Eight months on, it is the gap between the results and the share price that has become the most interesting piece of this file. In May 2026 Nvidia published record quarterly revenue of 81.6 billion dollars, up 85% year on year — and the stock fell the next day. Since its peak on 14 May 2026 the shares have lost around 18%. Three signals explain the gap: the collapse in GPU rental rates, with the B200 going from around 6.11 to 4.22 dollars an hour between late May and late June 2026, a 31% fall in three weeks; data centre revenue in China down to virtually zero under export controls; and sustained insider selling over the quarter.

My reading is this: the market has stopped paying for growth and started paying for the durability of that growth. It is not a crash, it is multiple compression. Historically that is what precedes one — without it being a rule.

On the OpenAI side, the valuation went from 500 billion dollars in October 2025 to 852 billion in March 2026, on an annualised revenue run rate of around 24 billion. The IPO, once floated for 2026, has been pushed to 2027 — officially to aim at the 1,000 billion mark, in practice because the market now demands answers on real profitability. A valuation-to-revenue ratio of around 35 poses a simple question: what is being bought, a company or a hypothesis?

It is a “circular economy” described by analysts: the same capital circulating between giants, inflating bubbles with no concrete near-term profit. Michael Burry (famous for predicting the 2008 crisis) accuses Nvidia of feeding that bubble, with misdirected investment.

That diagnosis is no longer the preserve of short sellers. In August 2026 the European Central Bank published a post arguing that a correction in US technology valuations is likely, even on the assumption that AI delivers every one of its technical promises. It puts euro-area households’ direct exposure to the Magnificent Seven at around 440 billion euros, with a comparable amount for the insurance and pensions sector. Its most uncomfortable argument is not about the level of valuations but about the capacity to absorb a shock: unlike in 2000, central banks today have far less room to cut rates or support the economy.

The irony of the subject: Nvidia established itself precisely through extreme specialisation. In other words, even Nvidia illustrates the “specialist” logic. It did not win by being a generalist, but by becoming the best in one precise domain.

The gap between the sector’s revenue and the amounts committed to infrastructure has narrowed over the past year, but it remains an order of magnitude: spending on compute centres continues to grow faster than the revenue it generates, and the break-even point of the 2024-2026 investment wave is not expected before the end of the decade. It is precisely that time lag that makes the sector sensitive to the cost of capital — and so to the Fed’s decisions.

Unlike the dot-com bubble (built on web promises with no infrastructure), the AI bubble is held back by tangible constraints such as operating costs and energy.

The energy challenge: the sinew of a war that has already begun

Imagine that all the data centres on the planet today consume as much electricity as France and Spain combined. That is the reality of 2026, and the slope has steepened again since January: consumption rose 26% in twelve months. By 2030 it could more than double.

To grasp the scale: every time you ask ChatGPT a question it uses 2 to 6 times more energy than a simple Google search. Multiply that by billions of daily queries and you see the problem.

The real difficulty? The infrastructure is not keeping up. In the United States, experts forecast a shortfall in electrical capacity equivalent to 35-50 nuclear power stations by 2030. In Europe, building a data centre takes 2 to 3 years between permits, environmental studies and grid connection. An eternity when the technology moves in months.

For the more technical among you: in June 2026 Gartner put data centre consumption at 565 TWh for the current year, against 447 TWh in 2025, for a power draw of 133 GW against 105 GW a year earlier. The United States accounts for 204 TWh of that, including 68 TWh for AI-dedicated data centres alone. By 2030 the projection passes 1,200 TWh and 291 GW, with AI servers making up close to half the total. The 945 TWh I cited in January were therefore a low estimate.

In short, the exponential growth AI promises runs into an elementary equation: available electricity. Without a revolution in the way we produce and distribute energy, the promise of an infinitely scalable AI risks meeting a very real wall.

China builds reactors in months, the West in years: what that means for the AI bubble

Some countries are faring better. China, with its fast growth in solar and nuclear (including thorium, cheaper, less radioactive, and producing less waste than uranium), holds 25% of the world’s data centres against 45% for the United States. Less hampered by regulation, it builds power stations in months rather than years, securing its AI lead. In November 2025 it carried out the first thorium-to-uranium conversion in a molten salt reactor, with a target of 150 reactors by 2050.

Europe and the United States risk falling behind, and a sequence of events could be enough to test the “infinite AI growth” narrative — with, potentially, a brutal repricing of the most exposed stocks. Let me state the frame deliberately: I am not a trader and what follows is neither a prediction nor investment advice. It is an illustrative scenario, built on dated public catalysts that historically tend to cause market adjustments. My point is not “this will happen” but “here is how it could happen” if several signals stack up.

What my January scenario got wrong

In January I placed this sequence in February-March 2026: Nvidia’s results on 25 February, GTC in mid-March, a Fed meeting. None of the three triggered the break I described. Let me say it plainly: I got the calendar wrong, and that is worth writing down — a scenario that does not play out on the expected date does not become false, it becomes better informed.

What these eight months have shown is that the adjustment did not come through an event, but through erosion: Nvidia’s stock has lost 18% since May despite record results, which is a far slower and far less legible way to correct than a drop on a publication. And above all they have shifted the nature of the risk. In January, the limiting factor I identified was electricity. Today it is the geography of production, with the tensions around Taiwan in particular. So here is the calendar as I see it for the end of 2026.

The scenario, readjusted: fourth quarter 2026

• 27-28 October 2026: Fed meeting. No economic projections, so limited in itself, but it sets the tone before the November sequence. A more restrictive message raises the cost of capital, and tech valuations rest precisely on distant cash flows, so they are very sensitive to the discount rate.
• 17 November 2026: Nvidia’s fiscal Q3 2027 results (date confirmed, after close). The central appointment. The market will not be looking at revenue, which will probably be another record, but at two lines: whether GPU rental prices stabilise or not, and the visibility given on the rollout of the Vera Rubin platform. Cautious guidance on those two points would strike at the heart of the “AI trade”.
• 28 November 2026: local elections in Taiwan (the so-called “9-in-1” vote). Eleven days after Nvidia. These local elections are no longer local: Taiwan’s National Security Bureau recorded a 60% rise in inauthentic social media accounts between 2024 and 2025, and more than two million instances of disinformation in 2025 alone. A sequence of military or informational pressure around that date would put Taiwan risk back at the centre of the table — and with it the entire AI chip supply chain.
• 8-9 December 2026: Fed meeting with projections (the dot plot). The last of the quarter, and the only one to publish rate projections. In a market already jittery after November, that is the catalyst capable of turning a sector correction into a broader repricing.

Taken individually, each of these events is absorbable. What interests me is their concentration into six weeks and, above all, the fact that they are no longer independent of one another: a Taiwanese incident mechanically degrades Nvidia’s visibility, and a constrained Fed reduces the system’s capacity to absorb shocks. That is the definition of rising correlation — the moment when diversification stops protecting you.

And I will say it again, because the nuance matters: I am describing a possible chain of events, not a forecast. The most likely scenario is still probably the one of the last eight months — a slow erosion, with no identifiable break.

Today AI is no longer an isolated “tech” subject: it is a narrative pillar of the markets. And because I work with AI daily, I also see its constraints (costs, energy, infrastructure, maturity of use). Things that can, at certain moments, catch up with the financial storytelling.

The real breaking point is no longer energy, it is the Taiwan Strait

In January I concluded that AI’s wall would be electrical. I stand by the diagnosis, but I rank it differently. An electricity shortfall is a problem of pace: it slows things down. A supply chain break is a problem of existence: it stops them.

One figure sums the situation up: TSMC concentrates in Taiwan around 95% of the world’s advanced CoWoS 2.5D packaging capacity — the assembly step without which an AI accelerator simply does not exist. By the end of 2026 the foundry should reach 120,000 to 140,000 wafers a month on that process, against around 70,000 in 2025, and as early as late 2025 Nvidia had reserved between 50 and 60% of that 2026 capacity. TSMC’s US plants will not relieve the bottleneck before 2028-2029.

In other words: almost all the world’s AI compute for the next three years depends on an industrial process concentrated on an island roughly 180 kilometres off the Chinese coast. That is not a geopolitical opinion, it is a production fact.

Where do things actually stand in autumn 2026?

The summer sent contradictory signals, which is exactly what makes it hard to read. On the pressure side: a record 244 Chinese vessels recorded in Taiwanese waters in July. On the apparent easing side: 125 incursions into the air defence identification zone in August, a notably lower level. From Taipei: a supplementary budget of 240 billion Taiwan dollars voted for drones, and a referendum approved on 28 August opening the way to a return to nuclear power, which would reduce dependence on imported liquefied natural gas — and so the leverage an energy blockade would offer. From Beijing: fresh purges in the high command, with two generals removed.

My reading, as an observer and not as a geopolitical analyst: the front line has moved from the military to the cognitive. The most likely scenario for late 2026 is not an invasion, it is a large-scale influence campaign around the 28 November vote, supported by AI systems built for the purpose. That is no more reassuring: markets do not price an invasion, but they price uncertainty very badly. A simple sequence of tension would be enough to lift CoWoS concentration risk out of the footnote where it has been filed for two years.

There is an irony here I find hard to ignore: the most strategic infrastructure of this century rests on a single point of passage, and AI is simultaneously the tool being used to weigh on the election of the territory that houses it.

And what if Elon Musk were once again a step ahead…

A bold vision: data centres in orbit through Starlink V3, announced in October-November 2025.

The advantages: unlimited solar energy (closer to the Sun), complete autonomy, and zero terrestrial regulation. No waiting for permits, no energy shortages. Technically it rests on high-speed laser links for communication (latency ~20-30 ms against terrestrial).

But major challenges: launch costs (billions, via Starship), heat dissipation in the vacuum of space, and robotic maintenance.

In the medium term (2030+) specialists consider it credible. Musk sees it as an extension of Starlink for AI payloads. Bezos is following with Blue Origin, making it an AI space race.

Conclusion

After months of testing, analysing and observing AI day to day — in my work, my personal projects, even in conversation with people around me — I have to admit to a small disappointment. We were sold a dazzling “new era”, an upheaval comparable to humanity’s great ruptures. Yet when I look back, the real revolutions touched everyone, deeply.

The discovery of fire let humanity survive, cook food, keep warm; every individual benefited. The invention of the wheel and of stone-age tools transformed work and travel for all. The industrial revolution, with the steam engine, the railways and electricity, connected continents, created mass employment, urbanised societies and changed the daily lives of billions. More recently, the arrival of the internet, Google and smartphones democratised information, instant communication and services. My grandmother uses WhatsApp, my father looks everything up on Google, and whatever one thinks of it, children navigate tablets from a very young age.

AI, for now, does not have that universal impact. It remains above all a powerful tool for professionals, creatives and developers — people who already have a precise objective. In my own circle, as in the recent statistics, the general public tries it out of curiosity but has not yet made it part of daily life as an indispensable reflex.

That does not mean AI is useless — quite the opposite. It is already hugely beneficial and will be more so: more precise medical diagnosis to treat cancers and rare diseases, safer roads and cities, faster discovery of new drugs, help with the energy transition through predictive models. It will save lives, reduce inequality in access to knowledge and boost productivity.

My conviction is therefore this: AI is a major, beneficial, structuring advance… but it does not yet have the universal, immediate character of the great mass-market revolutions. Perhaps the real “AI revolution” will come when it can no longer be seen: built in everywhere, effortlessly, the way web search has become. In the meantime, let us keep our clarity: we are at the start of a long trajectory, and that is precisely why it needs to be thought about, governed and used intelligently.

What eight months have changed in my analysis

In January I thought the AI market would fragment: it did, faster than I wrote. I thought specialisation would protect the niche players: Perplexity demonstrated the opposite, because a speciality a generalist can absorb is not a defensible position. I thought the main constraint would be energy: it is first industrial and geographic, and it sits in an assembly process concentrated on a single island. And I thought a correction would come through quarterly results: so far it has come through a slow erosion, with no trigger event.

I will revisit this at the end of the year, once the November-December sequence has passed. That is the principle of an article you keep up to date rather than one you archive.

Sources used:

• Jedha — comparison / market share (December 2025)
• Unite.AI — market share and growth summary (December 2025)
• Similarweb — traffic and share data (cited via Unite.AI)
• The Decoder — market share / Grok / trends (December 2025)
• FirstPageSage — market share / traffic (December 2025)
• Exploding Topics — growth and traffic indicators (December 2025)
• Grace Blakeley — Substack (October 2025)
• Yahoo Finance — analysis / Michael Burry / Nvidia (December 2025)
• The Atlantic — analysis (December 2025)
• Sequoia Capital — revenue against infrastructure investment estimates (2025)
• International Energy Agency (IEA) — data centre consumption / 2030 projections (2025)
• Goldman Sachs — electrical capacity shortfall / projections (2025)
• World Nuclear News — thorium / molten salt reactors (November 2025)
• NucNet — thorium / nuclear (November 2025)
• PowerMag — nuclear / projections (December 2025)
• Forbes — data centres in orbit / technical constraints (December 2025)
• SingularityHub — space data centres / limits (December 2025)
• NY Post — the AI space race (December 2025)
• Space.com — space race / feasibility (December 2025)

Sources added for the September 2026 update

• First Page Sage — AI chatbot market share (September 2026)
• Sensor Tower — State of AI Report 2026, mobile market share (May 2026, cited via TechCrunch)
• TIKR — Nvidia analysis: share performance, GPU rental rates, China exposure (June 2026)
• Wall Street Horizon — Nvidia earnings calendar (2026)
• US Federal Reserve — 2026 FOMC meeting calendar
• European Central Bank — post on the expected correction in tech valuations (August 2026)
• Gartner — data centre electricity consumption, 565 TWh and 133 GW (June 2026)
• American Enterprise Institute — China & Taiwan Update (1 September 2026)
• Global Taiwan Institute — Taiwanese local elections and cognitive warfare (May 2026)
• Taipei Times — Taiwan’s vote set for 28 November 2026
• Sector analysis of TSMC’s advanced CoWoS / SoIC packaging (August 2026)
• French business press — OpenAI’s valuation and IPO delay (2026)

Written by

Nicolas Peter – Web performance and SEO expert in Perpignan and Paris
Oli-via-net agency

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

Setting up reliable conversion tracking is essential if you want to steer your Google Ads campaigns properly.
Without tracking, there is no way to know whether your ads are actually generating sales.

In this article I explain how to configure Google Ads conversion tracking in under 15 minutes, on an e-commerce site, using Google Tag Manager.
The method shown here is simple, quick and perfectly suited to WooCommerce stores (and it transfers to Shopify too).

What this tutorial sets out to do

1. track every sale made on your e-commerce site
2. feed those sales back into Google Ads
3. measure the real performance of your advertising campaigns

⚠️ Important
This tutorial lets you count sales, but it does not yet let you calculate ROAS precisely (return on ad spend).
For that you will need an advanced e-commerce dataLayer, which will be the subject of a future tutorial.

The example used here

To illustrate this guide I am using the Passi’Jardin website, an online shop built on WordPress with WooCommerce.
The client sells Christmas trees online and wants to know how many sales actually come from their Google Ads campaigns.

Step 2 – Create the conversion action in Google Ads

Log in to your Google Ads account, then:

• Go to Goals
• Click on Summary
• Find or create the conversion action matching the purchase

In the example, the conversion starts with a specific ID (“94…”, say).
As long as Google Tag Manager is not installed, the conversion shows as needs attention, which is perfectly normal.

Step 3 – Create the conversion tag in Google Tag Manager

Now go into Google Tag Manager:

What to do:

• Click on New tag
• Choose Tag configuration
• Select Google Ads – Conversion Tracking

You will need to fill in two essential fields:

• Conversion ID
• Conversion Label

To find them:

• Go back into Google Ads
• Open the conversion action concerned
• Click on Use Google Tag Manager
• Copy the two identifiers

Then paste them into the GTM tag configuration.

Once they are filled in, Google Tag Manager automatically recognises the associated Google Ads conversion.

Step 4 – Configure the conversion trigger

A conversion should only fire when a purchase has genuinely been completed.

To do that:

Together these confirm, 100%, that a purchase has been made.

Step 5 – Test with a real purchase or a discount code

The best solution is still to test a real purchase. To avoid paying for nothing, though, it is worth creating a 100% discount code.

Enabling discount codes in WooCommerce:

• Go to WooCommerce settings
• Enable the option Enable coupons

Go to the Marketing tab, then click Add coupon:

• Percentage discount: 100%
• Free shipping (if available)
• Usage limit (1 to 3 tests)

If your site has no free shipping (so you do not pay delivery costs during the test):

• Add a shipping zone
• Configure free shipping
• Tie it to a specific postcode (to be used for your tests)

Step 6 – Identify the order confirmation URL

Place a test order with:

• the discount code
• the postcode you configured

Once the order is confirmed, look at the page URL. The key elements to note are:

• order / order-received
• the part of the URL after the ? (the order key)

Step 7 – Finish the trigger in GTM

In Google Tag Manager, configure the trigger with two conditions:

• Page URL contains order-received
• Page URL contains key=

This double condition avoids false positives and guarantees that the conversion only fires on a genuine order.

Step 8 – Test with Preview mode

Before publishing:

• Click on Preview in GTM
• Place another test order
• Open Tag Assistant

You should see:

• the Google Ads tag fired
• both URL conditions met

If so, the tracking is working correctly.

Step 9 – Publish and check in Google Ads

Once the test passes:

• Click Submit in Google Tag Manager
• Publish the configuration

All that is left is to generate sales and check that they come through in Google Ads.

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

Update of 22 October 2026. The GAIA method now has its complete guide, with its four principles, its dated practices and a self-assessment scorecard. Read the GAIA method, the complete guide to GEO.

The GAIA method (Generative, Authority, Information, Agents)

Published 3 December 2025 — updated 16 October 2026: self-contained passages, off-site reputation and AI crawler access. Update of 3 September 2026: the roll-out of AI Overviews and AI Mode in France, 2026 zero-click data, agentic protocols (MCP, A2A, AP2), and a correction on the energy cost of AI queries.

No, SEO is not dead… but old-school SEO is

For more than twenty years, organic search ran on a fairly linear model: find keywords, write content, earn backlinks, aim for the top 10 and collect the traffic.

For several months now, that model has been cracking in every direction.

Four major shifts are behind it:

1. Google’s AI Overviews, which answer directly in place of the sites (source: blog.google)
2. The explosion of zero-click searches, where the user gets their answer without ever leaving the results page (source: Similarweb)
3. The roll-out of AI summaries inside Google Discover, which rewrite articles in the Google app’s news feed (source: Search Engine Land)
4. The arrival of AI Overviews and AI Mode in France on 22 July 2026: long spared for neighbouring-rights reasons, the French market has now switched too (source: Blog du Modérateur).

The numbers speak for themselves:

• A Pew Research Center study published in July 2025 shows that when users see an AI summary, they click a traditional link on only 8% of visits, against 15% when there is no AI summary.
• According to the SparkToro/Similarweb study covering January–April 2026, 68% of Google searches now end without a single click, against 60% in 2024 and 49% in 2019 — up 12.5 points in two years, the sharpest acceleration of the decade.

In other words: organic traffic is moving towards AI answers, algorithmic feeds (Discover, News) and conversational interfaces (Gemini, ChatGPT, Perplexity). Search is not disappearing, but it is changing nature.

That is the context in which I designed the GAIA method.

What exactly is the GAIA method?

GAIA is a strategic method I formalised to think about search in the age of AI. The acronym stands for:

• G – Generative: optimise for generative search and AI summaries.
• A – Authority: build solid brand and author authority (E-E-A-T).
• I – Information: produce exclusive information that AI cannot easily replace.
• A – Agents: prepare your content for autonomous AI agents.

In French, Génératif, Autorité, Information, Agents.

This is not a “new Google rule” but a working method I use at Oli-via-net to help clients stay visible despite AI Overviews, Discover and the rise of zero-click.

1. G for Generative: aim for the citation rather than the click

With AI Overviews, Google uses site content to generate a synthetic answer, then displays a few links below that block. In the Pew Research study, only 1% of visits involve a click on a link inside the AI summary itself, and 8% on any link at all (summary plus classic results).

The consequence: fighting for a position alone no longer makes sense across a large share of informational queries.

The objective becomes twofold:

1. Be understandable by Google’s AI, and by the other engines.
2. Be reliable and structured enough to be picked as a cited source.

In practice, optimising the G in GAIA means:

• Structuring content in logical blocks: clear headings, hierarchical subheadings, short paragraphs.
• Adding FAQs with direct 40 to 60-word answers to the key questions.
• Using Schema.org structured data (Article, FAQPage, HowTo, Product, Service and so on).
• Covering complete topics rather than keyword fragments, so as to become a topical reference on the search intent.
• Writing paragraphs that stand on their own. An AI does not cite a whole page, it extracts a passage, so the figure, the source and the expert’s name belong in the sentence itself, not three paragraphs earlier.

People increasingly talk about GEO (Generative Engine Optimization): optimising no longer for the traditional search engine but for the generative engine that composes the answers. You will also see AEO (Answer Engine Optimization, the direct answer) and LLMO (the footprint left inside the models themselves): three angles on the same work, of which GEO is currently the most widely used term.

2. A for Authority: authority becomes the real currency of AI search

In an environment where AI synthesises information, the key question is:
“Which sources will the algorithm consider reliable enough to feed its answers?”

That is where E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) comes into its own.
The Google Search documentation is a reminder that, however content is produced — by AI or by a human — what counts is that it is useful, high quality and people-first, with signals of expertise and trustworthiness.

In parallel:

• Google Discover now displays AI summaries drawn from several sources, with the original articles cited as logos or secondary links.
• Publisher complaints and Similarweb studies show a sharp drop in traffic to news sites since the mass roll-out of AI Overviews and AI summaries (source: New York Post).

More than ever, authority signals are decisive:

• An identified author with demonstrated expertise (track record, client cases, talks, publications).
• Topics genuinely mastered and treated coherently across the site (topical authority).
• Brand reputation (mentions, citations, interviews, reviews, multi-channel presence).
• Consistent editorial quality, including on news content (Google News, Discover).
• What is said about the brand off its own site. AI engines also summarise reviews, forums, comparison sites and the press, and a recognised but poorly perceived brand is left out of their recommendations.

Since AI Overviews arrived, the authority of the entity — person plus brand — weighs more than an accumulation of isolated backlinks.

3. I for Information: produce what AI cannot generate on its own

Language models are excellent at rephrasing and summarising, but they remain dependent on their input data. They cannot invent:

• Real field experience from client engagements.
• Proprietary data from campaigns, audits and A/B tests.
• Fine-grained professional analysis, grounded in daily practice.
• Argued points of view built on concrete situations.

Recent best practice on AI-generated content converges on one point:
AI can help produce a first draft, but the value is created through human rewriting and enrichment — examples, nuance, use cases, specific angles.

In practice, for the “I” in GAIA, I recommend:

• Using AI as a structuring tool (outline, table of contents, title variants) and an accelerator.
• Reserving the final writing of sensitive passages — analysis, recommendations, case studies — for a human.
• Systematically documenting concrete results: curves, conversion rates, time or revenue gained.
• Signing content with a real author who stands behind it.

That kind of information remains hard for a general-purpose LLM to replace, and it is a durable competitive advantage.

4. A for Agents: prepare your content for autonomous AI agents

The last building block of the GAIA method covers a movement that is now well under way:
the shift from a web consumed by humans through a browser to a web consumed by AI agents — personal assistants, research agents, transactional agents.

At the end of 2025 this was an anticipation. In 2026 it is infrastructure:

• A protocol stack has stabilised: MCP (Model Context Protocol) to expose data and tools to a model, A2A for agent-to-agent communication, AP2 and x402 for agentic payment, WebMCP to make a page actionable by an agent, and Web Bot Auth to tell a legitimate agent from a scraper.
• Agentic browsers have left the lab: ChatGPT Atlas, Comet, Google’s AI Mode and the assistants built into mobile operating systems carry out complete tasks — searching, comparing, booking, buying — without the user opening a single one of your page templates.
• Agentic commerce is taking shape: the agent reads a product page, compares prices and triggers a payment. Whatever is not marked up does not exist for it.

To be usable by these agents, sites have to be:

• Structured: product data, services, prices and availability clearly marked up.
• Accessible: APIs, feeds or pages standardised enough for an AI to parse.
• Reliable: up-to-date, coherent information with no contradictions between pages.

Concretely, that means:

• Using Schema.org to describe services, products, reviews and FAQs.
• Organising content into reusable blocks (steps, procedures, lists).
• Thinking of the site as a machine-readable knowledge base, not only as a shop window.
• Letting AI crawlers in. The robots.txt explicitly allows their crawlers (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended), and key information remains readable without JavaScript.

A word on the llms.txt file, often presented as “robots.txt for AI”: Google has said it does not use it, and real adoption on the engine side remains marginal. Putting one in place costs almost nothing, but it replaces neither correct structured data, nor a page structure that is genuinely extractable, nor exclusive information. That is where the work happens.

With GAIA, this “Agents” dimension is not an optional futuristic layer but a strategic axis to build into editorial and technical design right now.

Conclusion: the GAIA method in one sentence

The method’s principles, practices and self-assessment scorecard are set out in the complete guide to the GAIA method.

Beyond visibility, the search landscape of 2026 also raises an energy question — one that is often framed badly. The “factor of 10” between a generative query and a classic search comes from a 2023 estimate (around 3 Wh per query) that has since been corrected: Epoch AI puts a ChatGPT query at roughly 0.3 Wh, and Google measures 0.24 Wh for Gemini in production, against 0.04 to 0.3 Wh for a Google search depending on the source. The real gap is therefore a factor of 2 to 3, not 10 — and the real subject is not the individual query but the aggregate volume. Optimising content, structuring information and improving site performance also means doing the groundwork for AI, reducing its energy footprint and gaining digital efficiency.

The GAIA method (Generative, Authority, Information, Agents) starts from one simple observation:

SEO in 2026 is no longer about “writing for Google”, but about
“structuring, proving and spreading expertise in an ecosystem dominated by AI”.

Generative: make sure search AI can rely on your content.
Authority: build a brand and authors that algorithms can legitimately choose.
Information: produce data and field experience that models cannot guess.
Agents: prepare your site for a future in which AI systems talk to each other, with your content as the reference.

Written by

Nicolas Peter – Web Performance and SEO expert in Perpignan and Paris
Oli-via-net agency

This is the basis on which I now support businesses that want to stay visible in a world where answers are increasingly generated, summarised and filtered by AI. To go further, I also publish a ranking of the best GEO experts in France, a ranking of the best GEO agencies and a ranking of the best SEO consultants, with the scoring methodology set out clearly.

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit

FAQ — SEO, GEO and the GAIA method in 2026

Is SEO dead in 2026?

No. SEO as it used to be practised — keywords, backlinks, top 10 — is yielding less, but search has never been more active. What changes is the target of the work: you no longer optimise only to rank, but to be understood, chosen and cited by generative engines.

What is the GAIA method?

GAIA is a working method formalised by Nicolas Peter (Oli-via-net) around four pillars: Generative (being citable by search AI), Authority (building entity authority), Information (producing what a model cannot invent) and Agents (preparing your site for autonomous agents).

What is the difference between SEO, GEO, AEO and LLMO?

SEO targets ranking in classic results. GEO targets the citation inside generated answers. AEO targets the direct answer and featured snippets, LLMO the footprint left inside the models themselves. In practice the last three describe closely related angles on the same citability work.

How do I know whether an AI is citing my site?

There is no Search Console for generative engines yet. The workable method remains manual sampling: run a stable set of queries on ChatGPT, Gemini, Perplexity and Copilot at fixed intervals, and record the sources cited. On the server side, the logs for GPTBot, ClaudeBot and PerplexityBot tell you who is reading you.

Should I create an llms.txt file?

It is not a priority. Google has said it does not use it and adoption on the engine side remains marginal. Putting one in place costs almost nothing, but it replaces neither correct structured data, nor an extractable page structure, nor genuinely exclusive information.

A website’s load speed is no longer a simple matter of comfort: today it is a decisive factor for organic search (SEO), user experience and conversion. Yet many WordPress sites using the Divi theme still suffer from poor performance, particularly on mobile.

In this article I explain how to improve your Google PageSpeed Insights score effectively, from a real case, with a clear, concrete method you can apply straight away.

Why Google PageSpeed Insights has become unavoidable

The Google PageSpeed Insights tool assesses a website’s performance against several technical criteria that Google uses to rank pages in its search results.

In particular it analyses:

• Real load speed
• The Core Web Vitals (LCP, CLS, INP)
• Behaviour on mobile
• JavaScript and CSS blocking
• Image optimisation

A poor score has direct consequences:

• A fall in SEO traffic
• A higher bounce rate
• Lost conversions
• Damage to the brand image

Why WordPress sites with Divi are often slow

The CMS WordPress is extremely powerful and flexible, but that richness can become a drag if the site is not properly optimised. The Divi theme is highly visual but also heavy by nature:

• Heavy script loading
• Multiple animations
• Complex modules
• Large stylesheets

Without advanced configuration, a Divi site can easily score below 50 on PageSpeed Insights, especially on mobile.

Step 1: Technical clean-up of the WordPress site

Before even talking about caching or images, you have to:

• Remove unnecessary plugins
• Clean up the media library
• Delete unused themes
• Check for scripts loading for no reason
• Back your site up through FTP + database, or with the free Updraft Plus plugin

For that you can use the free plugin Media Hygiene. Every active plugin adds to server response time and browser rendering.

The aim: lighten the foundations before optimising performance.

Step 2: Image optimisation (the biggest performance gain)

Images often represent 60 to 80% of a web page’s total weight. Imagify offers a free tier that lets you optimise 20 MB of images a month (roughly 200 images)

What to do:

• Automatic compression
• Conversion to WebP
• Resizing to the exact dimensions
• Lazy loading for off-screen images

The direct result:

• A massive gain on Largest Contentful Paint (LCP)
• A better mobile score first and foremost

Step 3: Caching and preloading resources

One of the best-performing caching systems with WordPress and Divi is WP Rocket, which costs around €30 a year. You can use the free alternative WP Fastest Cache, which will be a little less effective. A good cache configuration lets you:

• Cut load time dramatically
• Reduce server load
• Speed up display for returning visitors

The key elements:

• Browser cache
• Server cache
• Font preloading
• CSS / JS minification

On WordPress this step is absolutely essential if you want scores above 80-90.

Step 4: Divi-specific performance settings

Divi includes several built-in optimisation options that are often overlooked:

• Dynamic module loading
• Deferred JavaScript loading
• Asynchronous CSS generation
• Turning off unnecessary animations

These settings reduce the weight of the DOM and significantly improve visual rendering.

Step 5: Checking the Core Web Vitals

After optimisation we checked the following indicators:

• LCP (Largest Contentful Paint) < 2 s ✅
• CLS (Cumulative Layout Shift) < 0.1 ✅
• INP (Interaction to Next Paint) < 200 ms ✅

These metrics are now built directly into Google’s algorithm.

The results after optimisation

After the technical optimisation of the WordPress site using Divi, we ran another PageSpeed Insights test in Desktop mode. The results are markedly better:

• Desktop performance score: 97
• First Contentful Paint (FCP): 0.8 s
• Largest Contentful Paint (LCP): 1.2 s
• Total Blocking Time (TBT): 0 ms
• Speed Index: 0.8 s
• Cumulative Layout Shift (CLS): 0.008
• SEO: 100 / 100

In concrete terms, that means:

• The browser no longer spends time blocked by JavaScript (TBT at 0 ms)
• The main content appears very quickly
• The page is visually stable while it loads
• The site fully meets Google’s expectations on desktop performance.

The direct impact on organic search (SEO)

A faster site means:

• Better indexing
• Better positions in the results
• A better click-through rate
• More time spent on the site
• A better conversion rate

Technical performance is now a pillar of modern SEO.

Should you aim for 100/100 on PageSpeed Insights?

Reaching 100/100 is possible in some cases, but it is not always necessary or cost-effective. The real objective is:

• To be above the Core Web Vitals thresholds
• To offer a smooth mobile experience
• To avoid algorithmic penalties

A score between 85 and 95 is already excellent in 95% of projects.

An example, the score of our agency’s own site, Oli-via-net.fr

Free SEO & Digital Marketing Audit

Get a personalised analysis of your website in under 48 hours.

• A clear diagnosis (technical, content, visibility).
• The 3 priority actions to improve your results.
• Recommendations suited to your sector and your budget.

100% free, no commitment. Our aim is to show you how to generate more qualified traffic, more leads and more sales.

Name*

Email*

Phone*

Your website URL (optional)

What you need*

Google SEO

Visibility in AI answers (GEO)

Paid advertising (SEA)

Reviews and online reputation

Website design or rebuild

Automation and AI agents

Not sure yet

Your message (optional)

I agree that my data may be used to handle my request. It is never shared or sold, and I can ask to access or delete it at any time (GDPR).*

I agree

Email

Get my free audit