Google AI Overviews in France: How to Get Cited

Google launched AI Overviews in France on July 22, 2026, and clicks dropped. Selection criteria and 7 actions to be cited as a source.

Updated 12 min read

Contents
  1. What is an AI Overview, exactly?
  2. AI Overviews in France: what changed on July 22, 2026
  3. Why it's different from appearing in ChatGPT or Perplexity
  4. The 6 selection criteria for AI Overviews
  5. 7 concrete actions to get selected
  6. Best query types for AI Overviews
  7. How to measure your presence in AI Overviews
  8. Frequently asked questions

Since July 22, 2026, when you type a question into Google in France, an AI-generated answer can appear above the traditional results. That's the AI Overview. Google launched AI Mode, a conversational search interface, the same day (Abondance). The effect on clicks was immediate: across 963 French domains, Ahrefs measured, nine days after launch, a median click-through rate drop of 5.7%, and 23.1% for the most exposed sites (more than 20% of their queries showing an AI Overview). If your site isn't the source cited in that answer, you're losing traffic even when you're ranking well.

The good news: AI Overviews don't come from thin air. Google generates them from existing web pages — and in July 2025, Ahrefs found that 76% of cited pages ranked in the organic top 10. Its March 2026 update puts that figure at just 38%, as Google increasingly pulls sources from related (fan-out) queries. In other words, if you're already ranking well, you have an advantage, but it's no longer a guarantee. And SEO alone isn't enough: Google's AI selects sources based on specific criteria that this article breaks down.

What is an AI Overview, exactly?

The AI Overview (formerly known as SGE — Search Generative Experience) is a block of AI-generated text that appears at the top of certain search results pages. It synthesizes information from multiple sources to directly answer the user's question.

Unlike featured snippets that extract a passage from a single page, the AI Overview combines and rephrases information from 3 to 5 sources. It cites those sources with links, but the user gets their answer without clicking — which explains the drop in organic CTR.

AI Overviews appear primarily on informational and comparative queries: "how to choose a CRM," "difference between LLC and S-Corp," "best web hosting 2026." They don't appear (or rarely) on direct transactional queries ("buy iPhone 16") or navigational queries ("Facebook login").

AI Overviews in France: what changed on July 22, 2026

AI Overviews had been live in many countries for two years when they reached France. The delay stems from the dispute between Google and press publishers over neighboring rights (Abondance). Since July 22, 2026, AI Overviews and AI Mode have been available in French, rolling out progressively on mobile and desktop. According to Blog du Modérateur, Google lets publishers opt out of AI Overviews without any ranking penalty. AI Mode, launched the same day, works on a different logic: see our article AI Mode vs AI Overviews, what changes in France.

The first French measurements confirm what was seen elsewhere. Ahrefs' study (Search Console data from 963 domains, 28 days before launch compared with the 9 days after) found:

  • a median drop of 5.7% in CTR and 8.8% in clicks across the panel;
  • −23.1% CTR for sites where more than 20% of queries show an AI Overview;
  • nearly one site in two lost more than 10% of its daily clicks;
  • health was the hardest-hit sector: 22.2% of queries with an AI Overview, −20.4% CTR;
  • a near-linear relationship: each point of AI Overview rate costs roughly one point of CTR.

In the United States, where there is more hindsight, Seer Interactive (53 brands, January 2025 to February 2026) found that organic CTR on queries with an AI Overview fell from 3.19% to 1.31% in December 2025, before recovering to 2.36% in February 2026, versus 3.82% for queries without one. The gap does not go away over time. For the broader traffic picture, see why your Google traffic is dropping.

Why it's different from appearing in ChatGPT or Perplexity

Google's AI Overviews have one major distinction: they draw only from Google's index; a page must be indexed and eligible to be shown with a snippet to be cited. Ranking well helps without being required: according to Ahrefs, only 38% of cited pages are in the top 10. ChatGPT and Perplexity use their own crawlers (OAI-SearchBot, PerplexityBot) and several search sources, and can cite sources that Google doesn't rank on page one.

In practice, this means that for AI Overviews, traditional SEO remains the foundation: the better you rank for a query and its related queries, the better your chances of being cited in the corresponding AI Overview.

It's the GEO layer that then makes the difference: among the candidate pages, which one gets selected as the AI Overview source? That's where the structure of your answers, your evidence, and the freshness of your content come into play.

SEO vs GEO: what's different in 2026

The 6 selection criteria for AI Overviews

Google hasn't published selection criteria specific to AI Overviews. Its documentation even states that no special optimization is required to appear in them: no special file, no dedicated markup, the usual SEO best practices apply. The signals below come from observing results and from published studies.

1. Organic ranking position

This is the starting signal. AI Overviews draw primarily from results that Google has already judged relevant and high-quality. But according to Ahrefs, a growing share of citations now comes from pages ranking beyond the top 10, picked up through related queries.

This doesn't mean position 1 always gets picked. Google AI may select position 3 or 7 if it judges that page better answers the specific question. Being outside the top 10 lowers your chances, but it isn't disqualifying.

2. Direct, extractable answer

AI Overviews favor pages that contain a clear, direct answer to the question asked, ideally in the first few paragraphs.

If the user searches "how to improve your GEO score," Google AI will look for a page that starts by answering that question — not a page that talks at length about the history of GEO before getting to the answer.

The ideal format: a direct answer sentence, followed by a structured explanation with steps or key points.

3. Descriptive subheadings

Pages with H2/H3 tags that ask questions or clearly describe the content of each section are more easily parsed by Google's AI.

Effective example:

  • H2: "How to appear in AI Overviews"
  • H2: "What types of content get selected"
  • H2: "How long until you appear"

Ineffective example:

  • H2: "Our approach"
  • H2: "Our services"
  • H2: "Learn more"

4. Schema.org structured data

Accurate Schema.org markup (Article with its last-modified date, FAQPage matching the visible FAQ) helps Google read a page's content type, date, and structure. It isn't decisive, though: Google states that no special markup is needed for AI Overviews, and the Detekia calibration study of September 30, 2026 (317 pages from 40 sites, cited pages compared with non-cited pages from the same site) found that Organization schema did not distinguish cited pages. Your images and videos matter too (alt text, captions, transcripts): see multimodal GEO: images and videos.

The question-and-answer format matters most in the visible text: question headings followed by a short answer were what most clearly distinguished cited pages in that study.

Schema.org and AI: the practical guide

5. Brand and site reputation

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is often presented as a filter on sources. Yet Google points out that it is not itself a ranking factor, and in our study, an identified author did not distinguish cited pages from non-cited pages on the same site (an author bio remains useful to readers, and expected in health or finance). Reputation weighs more: across 75,000 brands, Ahrefs found that brand mentions on the web are the factor most correlated with visibility in AI Overviews (0.664), far ahead of backlinks (0.218).

This is where established sites have an advantage over newer ones. But an exceptionally well-structured and comprehensive piece of content on a smaller site can still get selected — especially on niche queries.

6. Content freshness

For queries where timeliness matters ("best CRM 2026," "new data privacy regulations"), Google AI favors recent content. Publication date and last-modified date are direct signals.

A 2024 article on an evolving topic will be passed over in favor of a 2026 article, even if the former has more backlinks.

7 concrete actions to get selected

Action 1 — Target featured snippets first

Featured snippets and AI Overviews often target the same queries. If you already have a featured snippet, you have a strong probability of being cited in the corresponding AI Overview.

Optimize your pages for featured snippets: concise answer in 40-60 words directly under the H2, followed by a detailed explanation.

Action 2 — Structure in "question → answer" format

Every important section of your page should follow this format:

  • An H2 or H3 phrased as a natural question
  • A direct answer paragraph (2-3 sentences)
  • A detailed expansion with examples or data

This format is exactly what Google's AI looks for when building its summaries.

Action 3 — Add lists and numbered steps

AI Overviews frequently display lists. If your content contains ordered lists ("5 steps to...," "the 3 criteria for..."), it has a better chance of being selected and displayed in the format users expect.

Action 4 — Cover the topic comprehensively

AI Overviews combine information from multiple sources. If your page covers the topic comprehensively — every facet, every common question — Google AI can cite it as the primary source rather than having to combine 4-5 partial sources.

Aim for 2,000+ word articles that address the topic in depth, with subsections for each aspect.

Action 5 — Add unique data points

AI Overviews frequently cite statistics and numbers. If your page contains original data (your own studies, benchmarks, measured results), it has an advantage over pages that merely compile data from other sources.

For example, an accounting firm publishing the average turnaround it observes on its own cases, or an online store publishing the actual return rate of its products, with the period and the measurement method. Original data is only valuable if it is dated and verifiable.

Action 6 — Keep your markup accurate, without expecting miracles

An Article schema with an accurate last-modified date and, if the page has a visible FAQ, FAQPage markup repeating the same questions are enough. They make the page easier to read, without being decisive: Google states that no particular markup is needed for AI Overviews. The time saved is better spent on the structure of your answers (action 2).

Action 7 — Update regularly

AI Overviews change frequently — Google regularly recalculates which sources to cite. Content updated with fresh data has a better chance of being selected than static content.

Update your pillar articles at least once per quarter. Add recent data, remove outdated information, and update the modification date in your Article schema.

Best query types for AI Overviews

Not all query types trigger AI Overviews. Focus your efforts on the ones that generate them most often.

"How" and "why" queries

"How to choose a...," "Why does my site...," "How does [X] work..." — these queries generate an AI Overview in the majority of cases. They call for an explanatory answer that AI can synthesize well.

Comparative queries

"X vs Y," "Difference between A and B," "Best X for Y" — comparisons are prime territory for AI Overviews because they require synthesizing multiple criteria.

Definition queries

"What is GEO," "Definition of [concept]" — Google's AI excels at synthesizing definitions from multiple sources.

Poor candidates

Navigational queries ("official site of X"), direct transactional queries ("buy X"), and hyper-local queries ("Italian restaurant nearby") rarely trigger AI Overviews.

How to measure your presence in AI Overviews

Google Search Console doesn't yet provide specific data on AI Overviews. But several approaches let you track your visibility.

Regular manual testing — identify your top 10-15 target queries. Check monthly whether an AI Overview appears and whether your site is cited in it. Use an incognito browser window to avoid result personalization.

CTR tracking in Search Console — if you see a sudden CTR drop on a query where you rank well, it's likely that an AI Overview appeared and is capturing clicks. If you're cited in the Overview, CTR may partially recover.

Third-party tools — tools like Semrush and Ahrefs are starting to track AI Overview presence in their keyword reports. Check whether your current SEO tools offer this feature.

GEO score — the selection criteria for AI Overviews overlap significantly with the 6 criteria of the Detekia GEO score, particularly answer structure, evidence and data, and freshness.

Multi-AI tracking — Google is only part of the picture. To measure your brand's presence in Gemini, ChatGPT, Claude and Perplexity, with results weighted by the actual usage of each AI in France, and Google AI Overviews collected on the same queries and shown separately, that is what Beeleven's AI visibility audit is for.

Frequently asked questions

Do AI Overviews appear on every Google search?

No. Their frequency varies across studies and query sets. According to Semrush (10M+ keywords), they triggered on 15.69% of queries in November 2025, after peaking at 24.61% in July. They mostly concern informational queries but are gradually expanding to commercial ones.

Can I prevent my content from appearing in AI Overviews?

Not with Google-Extended: that robots.txt token governs the use of your content for Gemini models, and Google states it has no impact on inclusion in Google Search. AI Overviews rely on Googlebot's index. To limit what gets reused, Google points to the nosnippet, data-nosnippet, max-snippet, or noindex directives, which also shrink your snippets in classic results. It's usually counterproductive: you lose a visibility channel. In France, Google also says publishers can opt out of AI Overviews without a ranking penalty (Blog du Modérateur).

Will AI Overviews kill organic traffic?

Not entirely. They redistribute traffic: sites cited in the Overview capture a share of it, others lose out. The goal is to be among the cited sources, not to fight the change.

My site ranks #1 on Google but isn't cited in the AI Overview. Why?

Ranking first in SEO doesn't guarantee being the AI Overview source. Other signals come into play: a short answer under each heading, question headings, comprehensiveness, freshness, brand reputation. A competitor in position 4 with better-structured content can be preferred.

Analyze your site — GEO score out of 100, 6 criteria

Test your GEO score for free

Free, no account, nothing to install. You get the result in exchange for your email.

The Detekia monthly letter

New articles and AI visibility news, one email per month.

Confirmation by email. Unsubscribe in one click from every letter. Privacy policy