Let's be honest: we've all, at some point, typed our brand name into Google just to see what comes up 👀 The new reflex is to do it in ChatGPT. And then, surprise: your brand is mentioned, or it doesn't exist. That's the whole point of... GEO (Generative Engine Optimization), the theme of our latest AI Red Thread, our quarterly event dedicated to artificial intelligence for communications professionals.
The real-world test: "Which communities or organizations offer training and information in communication in France?" posed to two AIs. The result: two very different answers. In one, We Are COM comes out on top. In the other, We Are COM doesn't even appear. Welcome to the world of GEO, where your visibility is no longer determined by a results page, but by a response written by a machine. Let's go 🚀
AI news you shouldn't miss (March to June 2026)
The Harvard Business Review publishes the 3rd edition of his study AI in the Wild, which lists the 100 main real-world uses of AI based on 12,637 analyzed cases. At the top: therapy and companionship (11% of uses, twice as much as a year ago!).
The study also coined the term "thinkslop": this lazy thinking that sets in when we delegate our thinking to machines. At work, the gains are real but remain individual; AI still rarely transforms processes in a profound way.
For the third consecutive year, the BDM outlines the landscape of generative AI tools used by digital professionals. And after two editions dominated by OpenAI, the market is now divided among several strong playersWe also note that the tools are... increasingly democratized. Text generation is almost universal, with 95,4% of users. Image generation follows (70,4%), ahead of coding tools (44,7%), then video, still a minority (24,2%).
Magic LayersCanva's technology, which breaks down a flat image into editable layers, is now integrated directly into ChatGPT and Gemini.
In practice: you generate a visual in the chat, you ask to transform it into a Canva design, and each element (text, logo, background, objects) becomes individually modifiable.
No more generating, downloading, re-uploading. And most importantly, no more near-perfect AI images that can't be corrected: the typography is adjusted, the color is aligned with the brand guidelines, and the format is adapted. 🎨
In June, Adobe deployed its AI Assistant in public beta in several of its software programsThe role of the assistant differs:
- Premiere : sorting rushes into bins, batch renaming clips, identifying interview questions or assembling a first edit.
- Photoshop: changing the background, resizing a visual for each platform or reorganizing layers, applied to the entire composition.
- Illustrator: producing files from a spreadsheet, reorganizing layers or checking to spot color mode errors and missing fonts.
- InDesign: applying a style guide from a PDF or template, updating the text, style and controls before printing.
Anthropic launched Claude Design to enable teams without design skills to create models, prototypes and presentations from a promptWhat really changes is that the tool can ingest your organization's design system and codebase to automatically apply your colors, typography, and components to each new projectAnd all of it exportable to CanvaPDF, PPTX or HTML.
At the end of May, Google made a series of groundbreaking announcements during its keynotes. Among them was the arrival of its new video generation model, promising even more impressive results. Gemini: Omni now generates videos from a simple text prompt, with a visual and narrative coherence bluffing. We are talking about sequences of several seconds, with transitions, soundscape, voice-over.
Google unveils an audio model for near-instantaneous voice translation in more than 70 languagesThe break: it translates continuously, while the person is speaking, preserving the speaker's intonation, rhythm and voice.
Deployed in Google Translate, in preview in Google Meet, and open to developers. International conferences, executive interviews, multilingual events: for our industry, it's simply mind-blowing. 🌍
On June 9th, Anthropic, the creator of Claude, released its two most powerful models, Mythos 5 and Fable 5, particularly adept at detecting computer vulnerabilities. Three days later, the US government ordered that access to them be cut off, in the name of national security, for all foreign nationals. Anthropic complied, deactivating the models for everyone, but publicly contested the decision, claiming it was a misunderstanding.
The stated reason: the fear that these models, capable of identifying vulnerabilities in almost all software, could be exploited by countries like China or Russia. Underlying this are long-standing tensions between Anthropic and the Trump administration.
Twelve days later, the models are still offline. Negotiations continue, Congress is demanding explanations, and a resolution to the crisis is expected sometime in July. It should be noted that the other models, including the one we use daily, are not affected.
And what this tells us, as communicators: a digital tool can be shut down overnight by a state decision. This clearly raises the question of our dependence on foreign technologies.
GEO, the new playground for visibility 🎯
GEO is All the techniques aimed at optimizing your content for generative AI search engines, so that your brand is selected and cited in their results.Where SEO aimed to climb in a list of links, GEO aims to exist. in the answer itself.
Why now ? Because habits are changingIt is estimated that 40% of French people already use AI as a substitute for a search engine, and that AI-generated traffic has increased by approximately 30% in France in one year. And while this traffic remains modest, the trend is clear. Above all, where AI is directly integrated into the search engine, clicks disappear: in the United States, nearly 70% of Google searches no longer result in a click. And knowing that when An AI responds, it only cites 2 to 7 sources…. Either you are one of them, or you don't exist for this internet user.
Good news, especially for France: Google AI Overview, that summary generated by Gemini at the top of search results, is still not available here. The reason isn't technical but legal, related to neighboring press rights and the ongoing legal battle that has already cost Google €250 million. The tech giant prefers to wait for clarity rather than risk another fine. For us, it's a real window of opportunity: we can prepare our AI visibility calmly, before the wave. And no, it's not a language issue; AI Overview already operates in French in Belgium and Switzerland. Rest assured, SEO isn't dead: GEO is its logical extension, built on the same foundation of authority.
How AI chooses who to quote (and why it's our job)
An AI doesn't reason like Google. It starts by breaking down your question into dozens of sub-queries, searches for reliable passages in its database and on the web, and then synthesizes everything. So we no longer think in terms of "my page," but rather "my ecosystem."
Which sources are most frequently cited? A trio dominates everywhere: Wikipedia, YouTube, and Reddit. And a newcomer is rapidly gaining ground: LinkedIn, now the fifth most cited source by ChatGPT according to the latest studies. This underscores the importance of taking the presence of your leaders and experts very seriously. Regarding sovereignty, let's keep in mind that AI is not neutral A study by the INA shows that in France, Nearly 26% of ChatGPT's outbound traffic to the press goes to a single media outlet, Le Monde., the first to have signed a partnership with OpenAI. On Claude's side, France Info is the most cited media outlet, with "only" 7,9% of outgoing traffic.
La 4th edition of the Havas AI brand tracker barometer, shows the evolution of the type of source of ChatGPT and Gemini, based on a panel of 150 questions representative of the most frequent queries of the French and taking care to adapt the distribution and diversity of the training to the specificities of each domain.
Conclusion: Things are really moving at ChatGPT! The share of Institutions has jumped by 30 points, reflecting its spectacular rise in legitimacy and a strategic shift towards offering users more reliable and credible information. The influence of Media outlets has decreased by 32 points (from 70% to 38%), marking a significant diversification of its sources, which reflects an active search to better meet the varied expectations of consumers. As for Institutions, they are establishing themselves at the forefront in the technical and sensitive sectors. At Gemini, the presence of Media outlets has increased by 11 points (from 29% to 40%), confirming its strong positioning in web content and thus consolidating an essential editorial profile for inspiration and quality information.
What makes a brand get acquired?
Several criteria determine whether an artificial intelligence will take over a brand. The first is authority and the reputationAn AI grants A brand cited by third-party sources carries far more weight than a brand that promotes itself.Being mentioned by a recognized media outlet, having a well-documented Wikipedia page, or seeing one's expertise relayed by credible third parties is infinitely better than the most eloquent statements published on one's own website.
6,5 times more chances of being cited via a third-party source (media, forums, comparison sites) than via one's own website
La freshness some content This constitutes the second criterion. Artificial intelligence values recent and regularly updated information. The "evergreen" content strategy, designed to remain valid indefinitely, shows its limitations here. Display a publication date, refresh your strategic pages and keeping its content alive over time becomes a true sign of reliability.
The third criterion is to name things with precisionExplicitly mentioning the names of your products, brand, or the people involved, rather than relying on vague wording, creates an anchor that AI can identify and leverage. The clearer and more consistent the vocabulary, the more recognizable the brand becomes to the models.
La content structureFinally, it plays a crucial role. Inverted pyramid writing, which places the essential answer in the first lines, FAQ-style formats, and short, easily extractable paragraphs are particularly appreciated by artificial intelligence, as they allow it to easily identify and reuse information.
One point of caution is necessary, however: artificial intelligences are not fooled. A study dedicated to B2B has demonstrated this Brands that present themselves as "the best" in their own articles see their page cited by AI, but in nearly seven out of ten cases, it's to recommend their competitors. The conclusion is self-evident: GEO is not a battle for visibility, it's a battle for credibility. And credibility is precisely what we do.
What we can activate starting today
First on his website
Verify that AI bots can properly read your pages, as they are sometimes blocked without anyone being aware of it—a point to investigate with your technical teams. Structure your data with clean markup, visible dates, and FAQ sections.
And especially, change your editorial approach We no longer write for a two-word keyword, but for the real question the internet user is asking. "AI communication training" becomes "What training is needed for a communications manager starting out in AI?" It's the great return of the long tail. Proprietary data—your studies, your barometers, your exclusive figures—is also particularly prized by AI.
The real question posed to AI is: "What training is needed for a communications manager starting out with AI?"
Before (SEO): a "AI Communication Training" page crammed with keywords.
Now (GEO): an article that directly answers, "For a communications manager starting out, aim for a 2-day training course covering prompts, tools and practical cases", with program, duration, prerequisites.
The real question is: "How much does communication training cost and is it eligible for funding?"
Previously: a vague mention "our rates upon request".
Now: a clear FAQ, "A day costs between X and Y €, financeable via OPCO and training budget", figures and schemes named.
The real question is: "Which should we choose, We Are COM or another organization?"
Previously: a self-centered "Why choose us" page.
Now: an honest comparison that positions us against others, formats, target audience, what makes us most relevant
Off its site next
And that's where communications professionals take back control. Press relations are once again becoming a major lever for AI visibility. Wikipedia and Wikidata tooEmployee advocacy and your leaders' LinkedIn presence are taking on a new dimension. And let's not forget customer reviews, which are closely monitored by platforms like Gemini. The foundational exercise can be summed up in one sentence: map out the questions your brand wants to answer, then identify the third-party sources that allow you to be featured there.
In summary…
💡 The We Are COM advice: Before any action plan, run a test. Ask three different AI platforms (ChatGPT, Gemini, Vibe, etc.) a question in which your brand should appear, and see if you're mentioned, with what information, and alongside whom. It's free, and it's the best initial audit.
And how do I measure all of that?
We are monitoring two indicators: the citation (am I mentioned?) and the Part de voice (For several search queries, what percentage do I represent compared to my competitors?). Tools exist to track your AI rankings, some very expensive, up to 600 euros per month, others available for free trial. Good news: since June, Google Analytics has featured a dedicated "AI Assistant" row for AI traffic.
Be careful though, a large part of the impact remains invisible according to SimilarwebNear 56% of AI-influenced traffic actually arrives via a traditional search.and a brand cited by ChatGPT is 2,5 times more likely to receive a visit within seven days.
And tomorrow? Three signs to watch for
Three developments to watch. AI Overviews will eventually arrive in France once the legal uncertainty is resolved, so it's best to prepare now. Agentic commerce and advertising in AI are becoming established: in the United States, OpenAI has already launched its Ads Manager, with paid links displayed below responses. And the framework is becoming stricter, with the European AI Act and its transparency obligation, applicable from August 2nd, which requires the disclosure of any content generated or manipulated by AI.
One thing is certain: GEO is not a break with SEO, it is its logical continuation, in a world where AI answers in place of the user. The window is open, and in France for a little longer than elsewhere: lasting visibility goes to the first entrants.
Have you ever typed your brand name into ChatGPT to see what it says? 👀
For further exploration...
Understanding GEO is one thing. Implementing it effectively within your brand and with your teams is quite another. That's precisely the focus of our masterclass. "The AI-driven search revolution: from SEO to GEO". In 1.5 hours, your teams leave with the keys to understanding this paradigm shift, anticipating the arrival of AI search engines and taking action now. A short and operational format, designed to collectively build expertise on the subject that will redefine brand visibility.
And because knowledge is shared among peers, Club members benefit from a 15% discount on all our masterclasses.
The replay of the AI thread and the presentation slides
