AI Overviews have been live in the United States since May 2024, and they now appear on roughly half of US Google searches (Google’s own estimate, reported by omnibound.ai in June 2026). France was the last major market to get them, on July 22, 2026: an Ahrefs study published August 18, 2026 measured a median 5.7% click-through rate drop across 963 French sites over the first nine days, and SE Ranking found 52.6% of French searches triggering an AI Overview. Visibility is no longer decided only in the list of results: it is decided inside the answer itself.
This guide gives you a complete method: understand what these engines look at, apply seven concrete actions, measure your presence for free or with a tool, and take stock with our interactive diagnostic.
The essentials
The rules of the game have changed. AI answer engines (ChatGPT, Perplexity, Gemini, Google’s AI Overviews) no longer return a list of links: they write the answer for you. For a brand, the question is no longer “am I ranking well?” but “am I cited, and with which sources?”.
The method in one sentence: build the identity, the content and the proof that these engines can verify and reuse, then measure your presence every month.
From ranking to citation: what has changed
An AI answer engine does not work like a classic search engine.
AI answer engine: definition An AI answer engine analyzes several sources, then writes a synthesized answer to a question asked in natural language, often with the sources it used. That is how ChatGPT (with integrated search), Perplexity, Gemini, Microsoft Copilot, Mistral Le Chat and Google’s AI Overviews work. Also see our Perplexity vs ChatGPT comparison.
For a brand, the consequence is twofold. First, fewer clicks: the user gets the answer without leaving the results page. Second, the answer itself becomes a reputational issue, because your prospects read what the AI says about you, not only what you say about yourself.
The growth curve is steep everywhere: in the US, AI Overviews went from 6.49% of queries in January 2025 to roughly half of queries fifteen months later (omnibound.ai, June 2026). The good news sits in the same data: in France, where AI Overviews just launched, about 90% of the domains cited already appear in Google’s organic top 50 (SE Ranking, August 19, 2026). SEO is not dying. It becomes the launch pad for your AI visibility. Google’s official AI Overviews documentation is explicit: no special action is required beyond standard SEO best practices. The conclusion that GEO extends rather than replaces SEO is ours.
Beware of “guaranteed” citations. No agency and no tool can guarantee you a place in ChatGPT’s answers or in AI Overviews: answers vary by query, by moment, by user and by session. Any offer along the lines of “your brand in ChatGPT in 30 days” should be read with caution. AI visibility is built on verifiable content and proof, not on a trick.
What AI answer engines really look at
To be cited, your brand must first exist in the sources these engines consult: the model’s training corpus (what it learned before its cutoff date) and the live indexed web (Google and Bing feed the search of ChatGPT, Perplexity or Copilot). A brand absent from the indexed web is absent from the answers. Period.
Four criteria make a brand get cited, and cited correctly:
Clarity
The AI must understand in one read who you are, what you do and who you serve.
An explicit “About” page, a named offer, a precise geographic area.
Consistency
The same information everywhere: name, description, address, activity.
When two sources contradict each other, the AI keeps the most repeated information, not necessarily yours.
Corroboration
Information confirmed by several independent sources carries more weight.
Customer reviews, press, directories, partners: third-party sources validate yours.
Proof
Figures, dates and quotations are the material the AI reuses.
“12,000 customers since 2019” is citable. “Recognized leader” is not.
This logic has been measured. In the academic GEO study (Generative Engine Optimization) presented at the KDD 2024 conference, researchers found that adding citations, statistics and expert quotations to a piece of content improved its visibility in generated answers, by up to 40% depending on the configurations tested. Practical translation: the more verifiable material your content provides, the more it gets reused.
The llms.txt file, useful or gadget? Google is explicit: no special file is required to appear in its AI search experiences. Asked in June 2026, John Mueller of Google called the usefulness of llms.txt “purely speculative for now”, in an answer reported by Search Engine Journal: the file has existed for years, and no AI system uses it. His core recommendation fits in three words: do not block agents. Check that your robots.txt does not exclude AI crawlers (GPTBot, PerplexityBot, Google-Extended) if you want to be cited.
The 7 concrete actions to get cited
- Lock down your brand identity. Write your name, your description, your address and your activity once, then use them identically everywhere. Start with your “About” page and your Google Business Profile, then align LinkedIn, directories and your social profiles.
- Answer the real questions your customers ask. List the questions you receive by email, by phone and in reviews, complete them with Google autocomplete and the “People also ask” section, then publish an FAQ and guides. Golden rule: the direct answer at the top of the page, the details after. If writing is a burden, an assistant like Nation AI helps you produce that content.
- Make your content citable. Dated figures, named sources, quantified examples, attributed quotations. Write each page so a model can extract one fact in one sentence. The KDD 2024 GEO study puts the effect at up to +40% visibility. And if you write with an AI, a good prompt generator saves valuable time.
- Build up third-party proof. Verified customer reviews, press articles, professional directories, partnerships, public appearances: these are the independent sources the AI cross-checks against yours. It is the slowest lever, and the most decisive.
- Take care of the technical signals. An up-to-date sitemap, structured data, fast pages, and a robots.txt that does not block the AI agents you want citing you. The llms.txt file remains optional: no major engine requires it.
- Occupy the platforms AI engines consult. The French example shows it: YouTube is cited in 51.3% of French AI Overviews, and social and video platforms take up half of the ten most cited domains (SE Ranking, August 19, 2026). An active YouTube channel or LinkedIn profile feeds the answer directly.
- Measure every month. Run your key queries on ChatGPT, Perplexity and Gemini, log mentions, tone and sources, and compare month over month. It is the trend that counts, never a single answer. The next chapter details the protocol.
Measuring your visibility in AI: the free protocol and the tools
Measurement is the step most competitors skip. Without it, you cannot know whether your efforts pay off. Two levels: the manual protocol, free of charge, and dedicated tools.
The manual protocol: one hour a month, zero dollars
- Define 8 to 10 queries that matter for your business: “what does [your brand] do?”, “is [your brand] reliable?”, “recommend 3 providers of [your trade] in [your city]”, “[your brand] or [your competitor]?”.
- Run them every month on ChatGPT, Perplexity and Gemini, preferably in private browsing to limit personalization of the answers.
- Log everything in a table: mentioned yes or no, tone of the answer (positive, neutral, negative), cited sources, cited competitors, date.
- Judge the trend, never a single answer: responses vary from session to session, even for the same query.
| Tested query | ChatGPT | Perplexity | Gemini | Trend |
|---|---|---|---|---|
| “what does [your brand] do?” | mentioned yes or no, tone, sources | mentioned yes or no, tone, sources | mentioned yes or no, tone, sources | compare each month |
Dedicated tools
To automate, tools query the AI engines every day and track your mentions, the associated sentiment and the sources used. Prices read on the vendors’ pricing pages on August 20, 2026: euro prices as published, dollar figures converted at the ECB reference rate of August 19, 2026. They change fast, check them before you commit. Other players exist (GetMint, MyRankingMetrics…), but their pricing pages were not accessible at the time of our review:
| Tool | What it tracks | Indicative price |
|---|---|---|
| Meteoria | prompt tracking across AI engines (French tool) | from 75 €/month (≈ 87 $) |
| Hikoo | GEO visibility platform (French tool) | from 69 €/month (≈ 80 $) |
| HubSpot AEO | 25 prompts tracked on ChatGPT, Perplexity and Gemini | 50 $/month (28-day free trial) |
| Qwairy | monitoring across the major AI engines | from 79 €/month (≈ 92 $), 65 €/month billed annually |
| Mentionable | daily tracking across 7 AI engines | from 79 €/month excl. VAT (≈ 92 $) |
| Semrush AI Visibility Toolkit | add-on on top of a Semrush subscription | 99 $/month per user |
Start free. The manual protocol is enough for the first months. Move to a paid tool once you already measure by hand and want to industrialize the tracking of dozens of queries.
Interactive diagnostic: where does your AI visibility stand?
Answer the three questions. Your score and your three priority actions appear instantly, no signup needed.
Express diagnostic of your AI visibility
Pick one answer per question. The diagnostic updates with every choice.
Your diagnostic
Score: 0 / 12
Answer the three questions: your diagnostic appears here.
The four levels, in detail
Starting out (0 to 3 points)
Your brand does not yet have a usable foundation: AI answer engines have almost nothing to cite, neither from you nor from third parties.
- Create the page that says who you are, what you do and who you serve.
- List the 20 questions your customers ask and publish structured answers.
- Open your Google Business Profile and your LinkedIn profile, with the same information.
In progress (4 to 7 points)
The basics exist, but the AI lacks verifiable material to cite.
- Turn your customer questions into an FAQ with the direct answer up top.
- Add dated figures and quantified examples to your main pages.
- Collect verified reviews and get third-party sources to cite your business.
Solid (8 to 10 points)
Your foundations are good: what remains is stronger external proof and a measurement routine.
- Go after press and partnerships: every third-party mention consolidates the answer.
- Launch the monthly protocol from the previous chapter and track your key queries.
- Occupy the platforms cited by AI Overviews, starting with YouTube.
Reference (11 to 12 points)
Your brand has what it takes to be cited. Your challenge becomes maintenance and precision.
- Keep things fresh: update your key pages and figures every quarter.
- Automate measurement with a dedicated tool to react faster.
- Document and iterate: note what moves your mentions and amplify it.
Summary: the 7 workstreams, in order
| Action | Effort | Impact | Start with |
|---|---|---|---|
| Consistent identity | Low | Strong | The “About” page and the Google profile |
| Useful FAQs and guides | Medium | Strong | The 20 questions from your customers |
| Dated figures and citations | Low | Strong | Reworking your 5 main pages |
| Third-party proof (reviews, press) | High | Very strong | Verified customer reviews |
| Technical signals | Low | Medium | robots.txt, sitemap, structured data |
| External platforms | Medium | Medium | A YouTube channel or an active LinkedIn profile |
| Monthly measurement | Low | Strong | The manual protocol above |
FAQ: your questions about visibility in AI answer engines
What is GEO?
GEO (Generative Engine Optimization) covers the practices that increase the probability that content gets used, cited or picked up by an AI answer engine. SEO targets a position in the list of results; GEO targets a place in the answer itself. The two reinforce each other: well-ranked pages are the most cited.
How do I get referenced on ChatGPT?
ChatGPT combines its training corpus with a live web search. Concretely: publish indexed, structured content, lock down your brand identity and accumulate third-party proof. Nobody can force a citation, but usable material raises the probability. For the general walkthrough, see our complete guide to ChatGPT.
Can ChatGPT or Gemini recommend my brand?
Yes, if credible sources describe your business consistently. The reverse is also true: if negative reviews and critical sources dominate, the AI may advise against you. An AI’s answer is the mirror of what is written about you: watch it every month.
Is the llms.txt file mandatory?
No. No major engine requires it, and Google calls its usefulness “purely speculative for now” (John Mueller, June 2026, in an answer reported by Search Engine Journal). The priority is elsewhere: do not block AI crawlers in your robots.txt.
Will SEO disappear with AI answer engines?
No. The data shows the opposite: in France, where AI Overviews just launched, about 60% of the pages cited come from Google’s organic top 10, and 90% of cited domains appear in the top 50 (SE Ranking, August 19, 2026). And Pew Research Center found US users still click through, just less often when an AI Overview is present. GEO extends SEO, it does not replace it.
How long before results show?
Count on several weeks to several months, with no guarantee: engines recrawl the web and update their corpora in waves. That is exactly why monthly measurement matters: it tells you whether the trend is going in the right direction.
AI visibility is built, not bought
AI answer engines are not a fad: in France, more than one search in two already triggers an AI Overview. The brands cited tomorrow will be those building today what these engines like to cite: a clear identity, useful and verifiable content, third-party proof and regular measurement. No trick replaces that foundation, and it is within reach of businesses of every size.
Your content is the fuel of your AI visibility
Nation AI does not rank your site for you, and no tool can guarantee you a citation. What it does is help you produce the structured content AI answer engines like to cite: FAQs, definitions, practical guides.
Or discover all of Nation’s AI tools.
Main sources: Google’s official AI Overviews documentation; omnibound.ai, AI Overviews statistics compilation (June 25, 2026); Pew Research Center behavioral study of AI Overviews (July 2025); Seer Interactive longitudinal CTR study (2025); Microsoft AI Economy Institute, Global AI Adoption in 2025; SE Ranking study on European AI Overviews coverage (August 19, 2026); Ahrefs study on the French launch (August 18, 2026); academic GEO study by Aggarwal et al., KDD 2024 conference (arXiv 2311.09735); Search Engine Journal, John Mueller’s answer on the llms.txt file (June 2, 2026). The figures quoted are dated at their publication: this market moves fast. This article is general information and does not replace personalized professional guidance.
