AI is starting to decide what to buy: your content must now convince humans… and agents
For twenty years, the digital journey looked like this:
need → Google → websites → comparison → purchase.
Then social media added a second route:
content → creator → recommendation → purchase.
A third path is now emerging:
need → AI → comparison → recommendation → action.
And it’s starting to deliver tangible results.
On 3 September 2026, the British retailer John Lewis reported that searches originating from AI agents now accounted for 2,5 %, as opposed to 0,3 % a year earlier. Its Managing Director Peter Ruis describes this progression as very rapid and affecting all age groups.
But the most interesting detail for creators and entrepreneurs is elsewhere.
John Lewis just opened a studio for influencers in its Oxford Street store to produce more content, the group considers that this material feeds both consumer attention and systems that now help them discover products.
In other words:
the content you publish today may no longer be used solely to convince a person.
It can also become part of the information used by an AI tomorrow when a user asks:
"What tool do you recommend? »
"What coach can help me? »
"What solution fits my need? »
« Where to book? »
This is probably one of the most important marketing developments to understand at the moment.
The change: users no longer necessarily visit ten websites themselves
Let's take a classic search. You're looking for a microphone to make facecam videos.
Yesterday you could:
- type your search in Google
- open five comparatives
- watch two YouTube videos
- read opinions
- open Amazon
- compare prices
Today you can start with:
"I make Reels facecams in a room that is not acoustically treated. Budget 150 €. I want a simple microphone to use on a smartphone. Give me three options and explain the differences. »
IA performs part of the synthesis work.
OpenAI says explicitly observe more and more users who start their shopping path in ChatGPT to explore, compare and decide what to buy. His Shopping experience now allows you to visualize and compare products directly in the conversation.
This reduces a fundamental element of the journey:
the number of sites the user visits personally before making a decision.
And that's where the subject becomes strategic.
SEO isn’t going away: a new layer is being added on top
We read more and more:
"The SEO is dead. »
Then:
"We need to do GEO. »
Then:
"No, now it's AEO. »
This battle of acronyms does not help companies much.
Google published in May 2026 a new documentation for sites that want to appear in its generative experiences.
His message is quite clear:
sEO fundamentals remain fundamental.
Google insists on:
- useful and original content
- technically accessible information
- local data, products, images and videos properly exposed
- the usual SEO good practices
- the importance of avoiding generic and interchangeable content
So the new strategy is not:
SEO OUT → IA IN.
It looks more like:
SEO + credible content + structured information + clear identity + updated data.
What AI agents are actually changing for a creator or a small business
Imagine a user asking tomorrow:
" I am an independent consultant at La Réunion and I am looking for a simple tool to centralize my links, present my services and allow reservation. »
The AI must identify several things. It must include:
- what solutions exist
- to whom they address
- what functions they propose
- where available
- their positioning
- their limits
- their tariffs
- and how to take action
If your site boils down to:
The ultimate solution to unlock your potential
humans already don't understand much. Neither is agent.
Compare with explicit wording
VIFLY is a platform for creators, coaches, consultants and independents. It allows you to centralize your links with a LinkHub, present your offers and manage online reservations with VIFLY Booking.
This sentence is less "creative". But it immediately responds to:
who + what + for whom + how.
And in a Web increasingly interpreted by machines, this clarity becomes an asset.
John Lewis is investing in influencers at the very moment when AI agents are making headway
This makes the John Lewis case particularly interesting.
In view of the progress of research by AI agents, the company is not only responding by:
"let's improve our product catalog. »
It also invests in more human content.
The brand has opened a studio in its Oxford Street store where influencers will be able to produce daily content, while developing its own video series.
Why?
Because an AI always needs information to reason. And this information comes in particular from:
- product pages
- official sites
- tests
- comparisons
- videos
- opinion
- media
- creative content
The paradox is therefore interesting:
the more users go through AIs, the more credible and original material produced can become important.
Creators can become sources, not just distribution channels
It is a major transformation of the creator economy. So far, a brand often looked at a creator as:
"an audience for rent.".
50 000 subscribers. 10 % commitment. A video. A campaign.
In an environment where agents synthesize the Web, a creator can also become:
a specialized source of information.
Let's take a picture chain. It publishes for three years:
- tests
- comparisons
- demonstrations
- workflows
- guides
When someone asks an AI:
« What purpose to choose to photograph a marriage with little light? »
this content is potentially a much more useful subject than:
"Here is a picture of the product. Buy now. »
The creator no longer has only:
the scope.
It also has:
publicly documented knowledge.
Generic content risks becoming even less valuable
Here's the other consequence. If an AI can easily produce:
"The 10 best marketing tools in 2026"
then publish a 501e almost identical article brings little value.
Google insists in its documentation 2026 on the interest of content non-commodity, i.e. not just interchangeable with a thousand other pages.
What becomes interesting is:
Real experience
"We used the tool for three months. »
Original data
"This is what we observed on 200 reservations. »
Methodology
" We compared the solutions against these five criteria. »
Expertise
"This is the situation we recommend A rather than B."
The grade
"This tool is excellent for X, but bad for Y."
This material is much more difficult to reproduce artificially.
And much more useful to a person or AI who is actually looking for an answer.
Your website must answer the questions that a customer service representative would normally resolve
Here is a good way to rethink your pages. Do not just ask:
" Is this page pretty? »
Ask:
“Can someone who does not know me understand exactly what I offer in thirty seconds?”
Then:
" Can a system extract important information without having to interpret my marketing jargon? »
For a performance, a good page should ideally explain clearly:
- to whom she addresses
- the problem addressed
- what is included
- operation
- important conditions
- the expected result without false promise
- how to buy or book
You'll notice something:
it's also exactly what a human wants to know.
Optimizing for AI should therefore not turn your site into text for robots. Most of the time, it should make it better for humans too.
Structured data is once again becoming particularly interesting
Schema.org structured data is not new. But they correspond perfectly to this new environment.
In particular, they make it possible to indicate explicitly:
this is an organization;
this is a product;
this is a local business;
this is an article;
this is an event.
Google explains that it uses this data to better understand the information on a page.
For an organization, Google recommends that information such as:
- name
- URL
- logo
- address where relevant
- telephone
- identity of the organisation
For a local business, LocalBusiness allows to communicate in particular time, address and other operational data.
Please note, however:
Schema.org is not a "appear in ChatGPT" button.
And a valid markup does not guarantee any Google rich results.
It is simply a proper way of expressing structured information.
The next battleground is the quality of your product information
Agent trade is particularly accelerating this trend.
Microsoft has for example launched a system ofcatalogue enrichment able to automatically extract and supplement product attributes to improve discovery and recommendation.
OpenAI extends its own Agentic Commerce Protocol discovery of products in ChatGPT.
Google develops theUniversal Commerce Protocol, presented as a brick of his future scenarios of agent trade.
The common message is clear:
the catalogue becomes a data intended to be understood by systems.
This means that:
« Premium T-shirt V2 »
is bad information.
Whereas:
“Men’s T-shirt in organic cotton, 220 g/m², oversized fit, black, sizes S to XXL”
gives much more matter. This requirement of precision will also gradually occur on services.
And what about a coach, consultant or therapist who doesn’t sell products?
Logic remains relevant. A user may request:
"Find me a consultant available this week who controls Google Ads for small businesses. »
Or:
" I'm looking for a coach specializing in X who offers video conferencing appointments. »
To be a credible option, your digital presence must be able to understand:
who you are, what you do, who, where, how to contact yourself, and what is the next step.
This is precisely where the organization of the digital presence becomes strategic.
Your LinkHub no longer needs to be just a collection of buttons
A VIFLY LinkHub may be particularly useful when responding to this clarification logic.
Step:
TikTok
YouTube
But rather:
Who I am
Working with me
My services
My resources
My best content
Book
The networks can come next. The goal is to transform the page into:
professional identity card + guidance system.
And this organization is useful to both humans and systems that will need to understand your activity.
Booking becomes interesting when AI moves from recommendation to action
This is where the business of agent trade becomes particularly interesting. The first step of the AIs was to:
answer.
The second is:
recommend.
The following is:
take action.
OpenAI already has scenarios where the user can go as far as checkout from ChatGPT in the United States with some compatible merchants and integrations.
Microsoft also develops Copilot Checkout to reconcile intent and transaction.
For a claimant, the natural equivalent is not necessarily the purchase of a product.
It's:
reservation.
The human journey can already become:
IA → discover your activity → profile → VIFLY Booking → niche → appointment.
VIFLY Booking does not today mean that a third party AI agent can automatically book in all scenarios.
It is mainly used to give the clear and structured final stage once the intention is created. And it's already very important.
Don’t hide your important information behind a confusing interface
Another principle. If your service only exists in:
- one image
- an inaccessible PDF
- a Story
- a video without transcript
- an interface requiring a connection
it becomes more difficult to discover and understand. A simple rule:
essential business information should exist in the form of HTML text accessible on your site.
For example:
- name of offer
- description
- cible
- prices when public
- conditions
- availability
- CTA
Visuals enrich. They should not be the sole location of the information.
Social content must also contribute to a sustainable future
Suppose you publish an excellent Reel:
"The three mistakes that cause reservations to be lost to consultants. »
It generates:
150 000 views. Then six months later, he disappeared from the flow.
If the idea is important, turn it into:
- article
- guide
- FAQ
- resource page
- case
Why?
Because you create an accessible asset that is sustainable. The social network creates the signal.
Your site builds memory.
This is probably where AI SEO needs to be understood
The wrong reasoning would be:
"How can we manipulate ChatGPT to quote me? »
The best:
" How can we become a source clear enough and useful to deserve consideration? »
This includes:
- clear identity
- expertise
- content
- evidence
- structured information
- update
- technical accessibility
The magic shortcut probably doesn't exist. And that's pretty good news.
Think ‘entity’, not just ‘keyword’
The classic SEO often asks you:
"On which keyword do you want to ranker? »
This question remains useful. But a second becomes important:
" What does an engine really understand about my business? »
For example:
Is it:
a blog?
a link tool?
saaS platform?
a reservation system?
The more consistent the information is between:
- site
- product pages
- profiles
- articles
- external information
the clearer the identity becomes.
Google precisely recommends using Organization to provide explicit information about a company and its online presence.
Create pages that drive real decisions
Excellent IA-ready content is not necessarily:
"What is coaching? »
It may be:
« Individual coaching or group training: what to choose ? »
Because the user is deciding. Other examples:
« Calendly or integrated reservation system: what differences? »
« Free link-in-bio or professional page: when to move on to a complete solution? »
"Freelance or agency: which model to choose for this project? »
These content facilitate comparison.
And the comparison is precisely one of the tasks that AI agents are gradually taking over.
Share your limits too
This may seem counter-intuitive. A company usually means:
"We are perfect for everyone. »
But a credible recommendation must be able to respond:
" In which case is this product not suitable? »
Content like:
VIFLY Booking is suitable if..
later:
VIFLY Booking is probably not suitable if..
can be much more useful than a page filled only with superlatives. Because he really helps decide.
And the response engines are just trying to do that.
Reviews and feedback remain powerful, but the evidence must be genuine
Systems can find:
- opinion
- returns
- forums
- videos
- discussions
It therefore becomes even more risky to artificially manufacture:
- testimonies
- false tests
- false advice
- false figures
The sustainable strategy is to build up real evidence.
Captures. Use case. Demonstrations. Experiences. Testimonials allowed. Contextualized results.
An isolated commercial claim has little weight. A claim supported by evidence has more evidence.
AI does not necessarily favour those who post the most
It is also a good reason not to turn VIFLY or any blog into IA content farm.
Google explicitly warns that generating large-scale pages without added value may be part of its policies against scaled content abuse.
It is better to produce:
2 excellent original resources
that:
40 generic items produced automatically.
Especially now that almost everyone can generate 40.
Rarity moves towards:
what you know AI doesn't already know.
Your benefit: producing data that others will use
This is the higher level.
Don't just comment on a trend. Create a data.
Example VIFLY:
" We analyzed 1 000 LinkHubs from creators: Here are the five CTAs that are most often placed in first position. »
Or:
"Analysis of 500] reservations: these are the three times when users give up. »
Provided, of course, that the data are real and properly anonymized.
Then you get something much more powerful:
information from which you are the primary source.
Journalists can take it back. The creators can cite it. Engines can find out. It is one of the strongest assets in a web filled with synthesises.
Content creation is therefore becoming even more strategic
One might think that the arrival of agents means:
"people will visit creators less. »
Maybe in some ways.
But John Lewis seems to be exactly the opposite bet.
As its research from AI agents progresses strongly, the brand invests in a studio that allows influencers to create more content.
Why?
Because AI does not invent commercial reality. It synthesizes signals that must come from somewhere.
The creator can become one of these sources.
Social media and SEO are becoming less distinct
For a long time:
SEO = Google.
Social = Instagram/TikTok.
PR = media.
Affiliate = sales.
Today, YouTube content can:
- generate views
- appear on Google
- be quoted elsewhere
- become a source for a user
- influence a recommendation
An interview can:
- building trust
- generate backlinks
- clarify expertise
The borders are disappearing. This is why a modern digital strategy needs to think in ecosystemnot in the channel.
The VIFLY method: Become Understandable → Credible → Recommendable → Actionable
This is the framework I'd remember.
1. Understandable
Your activity must be identifiable in seconds.
Who are you?
Who is it for?
What offer?
2. Credible
Add:
- expertise
- evidence
- practical cases
- original content
3. Recommended
Explain precisely:
- in which case your offer is adapted
- what she does
- its limits
4. Actionable
The person must know:
- where to buy
- where to book
- where to register
- where to contact you
That's exactly the VIFLY logic:
visibility → understanding → trust → action → opportunity.
Agent trade simply adds a new potential intermediary before visibility:
IA → visibility → understanding → trust → action.
What to do next? A 45-minute audit
This is the audit I advise today.
Step 1 ) ask three AIs to describe your activity
Test a neutral query.
Step:
« Explains VIFLY using lively.fr. »
But something close to a real user search.
Observe:
- are you found?
- are you correctly understood?
- what information is incorrect?
- which competitors appear?
Do not take the result as an "official ranking".
Use it as a diagnostic.
Step 2 ) Open your main page
Does an unknown person include:
who + what + for whom
in less than ten seconds?
Step 3 ) audit your offers
Each offer should have an explicit page.
Step 4 ) check your structured information
Organization. Breadcrumb. Product where relevant. LocalBusiness for a suitable local activity.
Step 5 ) find your generic content
Ask:
" Could 100 websites have published exactly the same thing? »
If so, enrich it.
Step 6 ) Create a real source
Test. Study. Comparison. Own data. Real case.
Step 7 ) Simplify the action
After the recommendation:
what does the user do?
LinkHub. Product. Contact. Booking.
There must be an obvious answer.
For creators: your niche must be clear
A profile that says:
Creator
is very difficult to qualify. A profile that says:
I help freelancers simplify their marketing and acquisition with digital tools and AI
is much more understandable.
Specialization helps:
- human
- the mark
- the engine
- potentially the agent
She doesn't lock you up. It makes you identifiable.
For artists: document your creative world
Same logic.
An artist must not turn his work into a technical file. But he can publish:
- approach
- influences
- technical
- works
- dates
- locations
- projects
- shop
An AI unable to understand who you are will have much more difficulty recommending you for a search of the type:
"meeting artist working around..."
The artistic identity remains human. The information around it can be structured.
For trainers: present the learning outcome
Avoid:
Processing premium training
Prefer:
Training of 6 hours for freelancers wishing to learn how to create and analyze their first Google Ads campaigns.
Then:
- prerequisites
- programme
- level
- format
- duration
- target result
A good training page almost looks like a card someone could pass on to a colleague.
This is exactly what an agent must also be able to understand.
For coaches and consultants: set out the context of the intervention
An AI cannot properly recommend:
"Coach X"
if no one knows what problems Coach X is working on.
Explain:
I intervene when..
This service is suitable if..
She probably doesn't fit if..
Then create a simple step towards making an appointment with VIFLY Booking when this corresponds to your activity.
Do I need to create a special file for ChatGPT?
Probably not the priority reflex. There are many recipes:
"Add this and all the AIs will quote you.".
Beware.
Google itself devotes part of its new documentation to deconstructing certain myths around the AEO/GEO and recalls that good SEO practices remain the basis of its generative experiences.
The priority must remain:
excellent website + excellent content + reliable information + correct structure.
Before the hacks.
Conclusion: tomorrow, your next client might never type your name into Google
He could ask:
"What solution fits my problem? »
And let an AI do the first sort. This behavior already exists.
John Lewis states that research from IA agents went to his home from 0,3 % to 2,5 % in one year.
OpenAI develops product discovery directly in ChatGPT. Google invests in Universal Commerce Protocol and in Agent shopping experiences. Microsoft also brings together agents, catalogues and checkout.
It would therefore be risky to prepare your activity only for a web where everyone still arrives through a traditional results page.
But the answer is not to produce content specifically "for robots".
It's almost the opposite.
Be so clear, specific, credible and useful that your activity can be properly understood regardless of the front door.
Your next step
Today, take your main page and ask yourself five questions:
Does an AI understand exactly what I propose?
Does a person understand who it is intended for?
Do I have enough evidence to support my assertions?
Are my essential information accessible and up-to-date?
After a recommendation, can the user act immediately?
If the last answer is no, structure your course.
Use your VIFLY LinkHub to centralize identity, content and offers.
And if your activity is based on appointments, use VIFLY Booking to turn an intention into a real niche.
Because in the coming Web, being visible may not be enough.
You'll have to be clear enough to understand,credible enough to be recommended and structured enough to allow action.
FAQ – Frequently Asked Questions
How to be visible in ChatGPT?
There is no method to guarantee a citation. The best strategy is to make your identity, offers and content accessible, accurate, original and verifiable on the Web, while applying good SEO and technical practices.
Is SEO still useful with IA engines?
Yes. Google explicitly states that SEO fundamentals remain relevant to its generative features and recommends, in particular, original and useful content as well as good technical accessibility.
What is Agent Commerce?
It is a model in which an AI agent can assist several stages of a purchase: search, comparison, recommendation and, in certain environments, transaction. OpenAI, Google and Microsoft are already developing infrastructure in this direction.
Do users already buy via AI?
Yes in some environments. OpenAI allows some checkout scenarios directly in ChatGPT in the United States via its Agentic Commerce Protocol and compatible merchants.
Do structured data ensure that an AI recommends me?
No. In particular, they allow engines to better understand certain information, but do not guarantee classification, quotation or recommendation.
Do we have to create hundreds of articles with AI to be more visible?
No. Google warns that the massive generation of content without added value can fall under its policies against scaled content abuse. Originality and value remain a priority.
Will IA agents replace Google?
It's too early to say. Behaviours are changing, but Search remains major and Google is integrating more and more generative and agentic features into its products.
SOURCES
OpenAI : , 24 march 2026.
Main official source for the transformation of the shopping path in ChatGPT. OpenAI indicates that more users start shopping in ChatGPT to explore, compare and decide what to buy, and extends its Agentic Commerce Protocol to product discovery.
Read the official announcement OpenAI
Reuters: John Lewis looks to harness AI agent shopping in 'difficult economy', 3 September 2026.
Fresh editorial trigger. John Lewis reports that research from IA agents has progressed to 2,5 % versus 0,3 % a year earlier and announces a studio dedicated to creating content with influencers in his Oxford Street flagship.
Google Search Central : , 15 may 2026.
Official source essential to avoid fake GEO/AEO shortcuts: Google recalls that its good SEO practices remain fundamental and insists on original content, useful and non-exchangeable.
Read the Google Search Central documentation
Google: Google Marketing Live 2026, 20 May 2026.
Official source on the evolution of the Google agent trade, the Universal Commerce Protocol and new experiences aimed at bringing together research, recommendation and commercial action.
View ads Google Marketing Live 2026
Microsoft: , 8 january 2026.
Official source on Copilot Checkout, Brand Agents and automatic catalog enrichment to improve discovery and recommendations.
Google Search Central : Organization structured data.
Official source for recommendations regarding an organization's identity, URL, logo and other administrative information that Google can use.