Key takeaways:
- An AI Overview cites your page when it answers a precise question in an extractable passage, sourced and signed by an identifiable expert.
- The French rollout is confirmed before 23 September 2026: the window to prepare your content is closing.
- Seven levers work: clean indexing, direct answer, Hn structure, named statistics, E-E-A-T, precise entities, brand authority.
- In May 2026, Google debunked several “AI hacks”: llms.txt, artificial chunking and “AI-specific” schema do nothing to earn a citation.
To be visible in Google AI Overview, your page must answer a precise question in an extractable passage, sourced and signed by a recognised expert. Everything else follows from that rule.
Does ChatGPT recommend your brand?
Measure your presence in AI engines and spot the competitors cited in your place.
No credit card, first scan in 5 min.
The shift is no longer theoretical. Organic click-through rate drops 61% on informational queries when an AI Overview appears (Seer Interactive, September 2025, across 3,119 queries). Ranking first on Google is no longer enough: you also need to be the source the AI cites in its summary. As an SEO consultant based in Nice since 2016, here is how to get there.
What is an AI Overview, and why aim for a citation in 2026?
An AI Overview is a synthetic answer block generated by Google’s Gemini model, displayed at the very top of the results page, above the classic organic links. Users get their answer without clicking, alongside a few source links shown as thumbnails.
Earning a citation matters because that block captures the click. By February 2026, AI Overviews appeared on nearly 48% of queries tracked by BrightEdge, up from 31% a year earlier. On affected queries, Ahrefs measured a 58% CTR drop in position 1 (December 2025, 300,000 keywords).
The nuance that changes everything: being cited protects your visibility. Brands cited in an AI Overview record a 35% higher organic CTR than uncited ones on the same queries (Seer Interactive, 2025). Not appearing accelerates traffic loss; appearing partly offsets it.
Google AI Overview lands in France: why the window is closing now
France was long absent from the rollout — not for technical reasons, but legal ones. The feature already runs in French in Belgium and Switzerland since March 2025. The blocker was neighbouring rights: since 2019, French law requires Google to compensate news publishers when it reuses their content, and AI Overview synthesises and rewrites that content.
The lock has now lifted. In a letter to French publishers on 29 June 2026 (revealed by Ouest-France, confirmed by Les Échos and Le Monde), Google confirmed the rollout of AI Overview and AI Mode in France before 23 September 2026, with three commitments: publisher opt-out, separate metrics for classic vs AI search, and continued compensation under neighbouring rights.
In practice, you have a few weeks to prepare your pages before the AI block settles into French results for good. Businesses structuring their content now will lead those who wait for activation. One field signal confirms it: on already-equipped markets (Belgium, Switzerland, Canada), Ahrefs measured in early July 2026 a 44% traffic drop for HubSpot across 554 keywords with strictly stable rankings — a pure click loss attributable to AI Overview alone.
How to get cited in an AI Overview: the 7 levers that work
Google put it in writing in its official May 2026 guide: optimising for generative search is still SEO. The levers below apply the fundamentals with more precision. They work together; none is enough on its own.
1. Make the page indexable and technically clean
Your page must first be indexed in classic search: an AI Overview draws from the same index as organic results. Allow crawling in robots.txt, ensure fast loading, correct mobile rendering and clear architecture. A page invisible to Googlebot is invisible to the AI.
2. Answer the question directly, right away
Put the full answer at the start of the section, not after three intro paragraphs. The AI extracts a passage that resolves a question on its own. Open each block with a direct statement, then expand. This inverted pyramid logic increases your extraction odds.
3. Structure with a readable Hn hierarchy
Use a clean H1 / H2 / H3 hierarchy where each heading announces an answer. Question-style headings (“How…”, “Why…”) match how users query the AI. Pages that win a featured snippet are also more likely to be reused in an AI Overview.
4. Add named statistics and sources
A verifiable, data-backed claim is extractable; a vague statement is not. Add precise figures attributed to a dated source (“according to Ahrefs, December 2025”). The foundational GEO study (Princeton, Georgia Tech, IIT Delhi, KDD 2024) shows that combining readability + adding statistics outperforms any isolated strategy by over 5.5%.
5. Demonstrate E-E-A-T through evidence
Experience, expertise, authoritativeness, trustworthiness weigh heavily in AI source selection. Sign your content with an identifiable author, a detailed bio and real references. Don’t claim expertise, prove it: concrete cases, proprietary data, transparent methodology.
6. Name entities systematically
AI systems build their answer from named entities. Writing “a rank-tracking tool” creates no anchor; writing “Google Search Console” or “Ahrefs” does. Explicitly name tools, brands, people and concepts (query fan-out, passage indexing, neighbouring rights).
7. Build brand authority and mentions
Mentions of your brand across the web matter more than ever. According to Position Digital (2026), brand mentions show the strongest correlation (0.664) with appearing in AI Overviews, versus only 0.218 for backlinks. Multiply legitimate citations of your name tied to your topic, on recognised platforms.
What Google officially debunked in May 2026 (do NOT do this)
On 15 May 2026, Google published its first official guide dedicated to generative search. It dismantles several “AI hacks” sold for eighteen months. Time spent on these tactics moves you no closer to a citation.
- The
llms.txtfile: Google does not read it to generate AI Overviews. It has no effect, positive or negative, on Google Search (15 June 2026 clarification). Keep it for other assistants if you like, but don’t pay for it. - Artificial chunking: no need to fragment content into single-idea micro-blocks. Google’s systems understand multiple topics on one page. A clear, natural structure is enough.
- “AI-specific” schema: no AI-dedicated structured data exists. Standard schema.org stays useful for identity and the Knowledge Graph, but bolting it on “for AI” without editorial quality yields no citation lift (Ahrefs study, 11 May 2026, 1,885 pages).
- Rewriting “for AI”: rewriting content specifically to please the AI is treated as low-value content. Write clearly for a human reader; the AI follows.
- Fake mentions: artificially inflating brand citations is spam and backfires.
The rule is healthy: there is no technical shortcut. Useful, accessible, trustworthy content remains the only foundation.
Which content and queries trigger the most AI Overviews?
AI Overviews don’t appear everywhere. They favour informational queries — “how”, “why”, “what is” — where a summary adds real value. Navigational queries (reaching a specific site), transactional ones (buying) and local ones trigger very few.
Query length also matters: per Ahrefs, 46.4% of queries with seven words or more trigger an AI Overview, versus 9.5% for single-word queries. And 76.3% of queries that trigger a summary are non-branded: hence the importance of covering the generic questions in your market.
| Query type / industry | AI Overview trigger rate |
|---|---|
| Informational queries | ≈ 88% of cases |
| Healthcare (medical YMYL) | ≈ 88% |
| Education | ≈ 83% |
| B2B Tech | ≈ 82% |
| Transactional / shopping queries | ≈ 3% |
| Local queries (“near me”, city + service) | ≈ 4 to 8% |
Sources: BrightEdge (February 2026), Ahrefs (November 2025), SE Ranking.
The strategic read: your informational content (guides, comparisons, definitions) is the most exposed — so the most rewarding to optimise for citation. Your transactional and local pages stay largely protected.
How to measure whether you’re cited (and track competitors)?
Google Search Console includes AI Overview data in the performance report (“Web” search type), but does not isolate it from other clicks. You see an overall trend, not your share of AI citations. For precise tracking, three approaches combine:
- Manual testing: type your target queries into Google and note whether an AI Overview appears and cites you. Simple, free, done weekly on priority queries.
- SEO tools’ SERP filters: Semrush, Ahrefs and SE Ranking offer dedicated AI Overview filters to track presence and competitors.
- A dedicated AI visibility tool: to measure your mentions across ChatGPT, Perplexity, Gemini and Google, your share of voice, and the competitors cited in your place.
FAQ: being visible in Google AI Overview
Do I need to rank well on Google to appear in an AI Overview?
Ranking well helps but isn’t enough. Cited sources aren’t always the top-ranked: a page ranking 2 to 10 can be reused if it better answers a sub-question. The page must, however, be indexed and eligible in classic search.
Is structured data (schema) mandatory?
No. Google confirmed in May 2026 that no “AI-specific” schema is required. Standard schema.org (Article, FAQPage) stays useful for identity and the Knowledge Graph, but it doesn’t buy a citation on its own.
Does an llms.txt file improve my AI Overview visibility?
No. Google doesn’t use this file for search: it has no effect on your AI Overview citations. It may hold minor value for some other assistants, nothing more.
When do AI Overviews arrive in France?
Google confirmed the rollout of AI Overview and AI Mode in France before 23 September 2026. The delay was due to press neighbouring rights, now framed by Google commitments.
How long before I get cited after optimising a page?
On an established domain, a well-optimised page can appear as a citation within a few weeks of publishing or updating. For a newer domain, expect several months of consistent effort before meaningful visibility.
Do AI Overviews spell the end of SEO?
No. Google still holds nearly 90% of the French search market, and more than half of queries show no AI Overview. SEO remains the foundation; GEO (Generative Engine Optimization) adds the “citation by AI” layer.
Sources
• Seer Interactive (September 2025) — CTR Study: AI Overviews impact on organic and paid CTR, 3,119 queries / 25M impressions.
• Ahrefs (December 2025) — AI Overviews Reduce Clicks, 300,000 keywords (update) — ahrefs.com
• BrightEdge (February 2026) — AI Overviews coverage across 9 industries (48% of tracked queries).
• Google Search Central (15 May 2026) — Optimizing your website for generative AI features on Google Search — developers.google.com
• Google Search Central (15 June 2026) — Clarifying guidance on llms.txt files.
• Aggarwal et al., Princeton / Georgia Tech / IIT Delhi (KDD 2024) — GEO: Generative Engine Optimization (arXiv 2311.09735).
• Ouest-France / Les Échos / Le Monde (29-30 June 2026) — AI Overview and AI Mode rollout in France.


