The change in 60 words
Since July 22, 2026, Google shows French users a generated answer above the results: AI Overviews. It does not appear on every search. Public data shows it mainly targets questions and comparisons, cites about eight sources, and your product page is only one of them. Here is what that changes for a Shopify store selling in France.
France status as of September 17, 2026. AI Overviews and AI Mode have been available in France since July 22, 2026, on mobile, desktop and the Google app, powered by Gemini 3.5 (Google France). The studies quoted below cover US queries: no public study of the same scale covers France yet. The trends are useful; the exact percentages do not transfer as is.
Worldwide, Google reported more than 2.5 billion monthly users for AI Overviews in May 2026 (Google I/O 2026). The question is no longer whether your shoppers will see them, but on which searches.
Finding 1: overviews answer questions, not keywords
The clearest signal comes from an academic study published on May 13, 2026 (Xu, Iqbal and Montgomery, arXiv). The authors tracked 55,393 trending queries over 40 days (March 13 to April 21, 2026), from the United States.
| Search type | Share with an AI Overview |
|---|---|
| All queries | 13.7% |
| Question queries | 64.7% |
| Non-question queries, 1 word | 9.9% |
| Non-question queries, 5 words | 18.5% |
| Non-question queries, 6+ words | 38.7% |
Source: arXiv 2605.14021, US queries, March to April 2026.
Ahrefs reached the same conclusion on a much larger but older dataset (146 million desktop results, September 2025): about 58% of question searches show an overview versus 15.5% of the rest, and 46.4% of searches with seven or more words do (Ahrefs, November 10, 2025).
Google confirms the usage shift in its French announcement: an AI Mode search is on average three times longer than a traditional search (Google France, July 22, 2026).
In practice, “firm mattress” is unlikely to trigger an overview. “Which firm mattress should a side sleeper with back pain choose” is far more likely to. Your product page needs the facts that answer the second one.
Finding 2: few overviews on purchase searches, more and more on comparison
Direct purchase searches rarely trigger an overview. Ahrefs measured in September 2025:
| Search intent | Share with an AI Overview |
|---|---|
| Informational | 21.4% |
| Commercial (comparison, reviews, “best”) | 4.3% |
| Transactional (“buy”, “price”) | 2.1% |
| Shopping queries | 3.2% |
Source: Ahrefs, 146 million desktop results, September 2025.
The recent trend points in a clear direction. Across more than 600,000 US keywords tracked from November 2025 to April 2026, Semrush found that the share of commercial results with an overview grew 71%, while transactional results fell 5% (Semrush, July 2, 2026).
For a merchant, this means the overview steps in before the shopper picks a store, while they compare products and criteria. That is where your product is kept or dropped.
Finding 3: complex-decision categories are the most exposed
Semrush measured the strongest growth in finance: the share of commercial results with an overview rose 231% in six months (Semrush, July 2026). The arXiv study, by contrast, notes markedly lower rates on politically sensitive topics.
What the exposed categories share: the shopper needs explanations, criteria and comparisons before choosing. In ecommerce, that covers every product that requires a trade-off: bedding, skincare, supplements, sports equipment, appliances, baby gear. The more advice your product needs, the more likely an overview sits between the shopper and your page.
Finding 4: your product page is one source among eight
The arXiv study analyzed 7,583 overviews. Three numbers matter for a store:
- A median of 8 sources per overview (range 1 to 32). When your page is cited, it shares the answer with third-party reviews, press, videos and comparison sites.
- Nearly 30% of cited domains do not appear on the first page of regular results. Ranking well is not enough, and not ranking does not rule out being cited.
- 11% of claims are not supported by the cited page, mostly through omission. An overview can attribute a feature to your product that your page does not state.
Two direct consequences:
- Consistency across sources matters more than before. If your page shows one price, your Merchant Center feed another and a comparison site a third, the overview has to decide, sometimes without you.
- Each engine has its own source hierarchy. Your position in ChatGPT or Perplexity does not predict your position in AI Overviews. Our GEO signals by engine matrix details the differences. Measure each engine separately.
What Google asks for, and what it does not
Google is explicit: there are no additional requirements to appear in AI Overviews or AI Mode, and no files or markup to create (Google Search Central). The basics remain:
- indexable, accessible pages;
- helpful, people-first content;
- up-to-date information, including Merchant Center and the Business Profile for businesses;
- tracking in Search Console, where AI feature traffic is counted under the “Web” search type.
Google also added a Search Console control to decide whether a site appears in these generative features, with no effect on regular ranking (Google France). The technical detail (JavaScript, structured data, llms.txt) is covered in our Google AI 2026 guide for Shopify.
What it changes for a Shopify product page
The four findings lead to five actions, in this order:
- Answer comparison questions on the page. Who the product is for, who it is not for, which criteria set it apart, how it differs from the neighboring model. A short, factual product FAQ beats a marketing description. See our guide to optimizing a Shopify product page for AI.
- Align price, stock and reviews across the page, structured data and Merchant Center. These are the facts the overview compares across sources. Shopify’s Google & YouTube channel syncs the catalog, but not inconsistencies introduced by a theme or an app.
- Make reviews and delivery times readable without JavaScript. In our Shopify GEO barometer of August 13, 2026, only 6.1% of the 457 product pages read exposed a customer rating and 4.6% a delivery time without running JavaScript. Google renders JavaScript, but other AI engines often do not, and reviews loaded by widgets stay invisible to them.
- Exist beyond your own site. With eight sources per answer, reviews on third-party platforms, specialist press and demo videos count. They should state the same facts as your page.
- Measure Google on its own. Track the “Web” search type in Search Console and regularly check what the overview says about your products, especially on comparison questions in your category. With 11% of claims unsupported by the cited page, a mistake about your product is a real risk; our article on brand confusion in AI Overviews details a test protocol.
Limits of this data
- US datasets. The arXiv, Ahrefs and Semrush studies cover US queries. Behavior in France may differ, not least because the launch there is recent. For stores selling in the US, see our analysis of AI Overviews in the US.
- Dates. The Ahrefs study dates from September 2025, before most 2026 changes; its orders of magnitude are confirmed by the March to April 2026 arXiv study, not its exact percentages.
- Small samples per category. The arXiv study includes 158 shopping queries (14.6% with an overview), too few to derive a reliable ecommerce rate.
- Tool vendor studies. Several tracking tool vendors share France-specific figures in their newsletters. Until the method and data are published, we do not repeat them.
Check what AI reads on your product pages
Verity Score is a Shopify GEO app: it checks what Google, ChatGPT, Claude and Perplexity can read on every product page, flags gaps between the page, structured data and Merchant Center, then writes the fixes you approve to your store. Free trial on the Shopify App Store.