Google AI Overviews: how to adapt your SEO strategy to stay visible?

Since the massive deployment of AI Overviews in the United States in May 2024 and their gradual extension in Europe in 2025, Google Search Console data from many e-commerce sites shows a drop of 15 to 35% in CTR on informational queries (« how to choose X », « what is Y »). But on transactional queries (“buy X”, “price of Y”), the impact is almost zero — even positive for the sites that appear in the citations.

What exactly is Google AI Overviews?

Google AI Overviews is a feature that generates a summary response at the top of search results for certain queries, citing multiple sources. This response is generated by an LLM (language model) powered by information on the web, particularly pages that Google considers authoritative on the subject.

The difference with featured snippets: the AI ​​Overview cites several sources simultaneously (generally 3 to 6 links) and synthesizes the information. A featured snippet only cited a single page. This new feature means that appearing in an AI Overview is technically more difficult than a featured snippet, but that several sites can benefit from it simultaneously.

AI Overviews and LLMs like ChatGPT or Perplexity share a similar logic for selecting their sources. This is why a GEO (Generative Engine Optimization) campaign simultaneously works on the presence in Google AI Overviews and in other generative AI. Our GEO LLM article details this approach.

What is the real impact of AI Overviews on the CTR of e-commerce?

The available data (SEMrush, BrightEdge, Conductor studies — 2024-2025) allows us to paint a nuanced picture:

• Informational queries (how, why, what): CTR drop of 20 to 40% for positions 1 to 5. AI Overviews respond directly without requiring a click.
• Commercial queries (comparisons, reviews, best): CTR drop of 10 to 20%. Users still click to validate the comparisons presented in the AI Overview.
• Transactional requests (purchase, price, delivery): almost zero impact, or even slightly positive for the sites mentioned. These queries trigger AI Overviews less often because Google prefers to display Google Shopping and direct links to merchants.

Practical conclusion for an e-commerce: if your organic traffic is mainly based on transactional requests, the impact of AI Overviews is limited. If you have an informational blog that generates 40 to 60% of your traffic on advice queries, the impact can be significant — and requires a strategy adaptation.

Which e-commerce pages are most impacted by AI Overviews?

In order of decreasing impact:

• General blog articles (how to choose, basic buying guide): highly impacted. The AI Overview often answers as well as the article.
• Generic product comparison pages: moderately impacted. Comparisons with real data and opinions hold up better.
• FAQ pages and short content pages: impacted if the content is factual and without depth of analysis.
• Category pages with editorial content: little impacted — Google prefers to display merchants for queries with purchase intent.
• Product sheets: very little impacted — AI Overviews do not replace the act of purchasing.

How to be cited in Google AI Overviews?

Google does not publish official selection criteria for AI Overviews citations. But analyzes of the cited pages reveal constant patterns:

• Schema.org structured data: pages with FAQPage, HowTo, Product schema are over-represented in citations. Google can easily extract and structure their content.
• Strong thematic authority: domains that cover a subject in depth (semantic cocoons, numerous related pages) are favored over sites that cover the subject on the surface.
• Demonstrated E-E-A-T: identified author, biography, expert quotes, real data — the signals of authentic human expertise.
• Original and differentiated content: pages that provide a unique perspective (proprietary data, real feedback, comparisons based on tests) are cited more often than generic content.

The most effective strategy for getting cited: produce content that answers specific questions better than anyone else, with figures, concrete examples and structured data. This is exactly what our semantic cocoon architecture builds — deep topical authority that promotes citations in AI.

What types of content are most resistant to AI Overviews?

Some content formats are structurally difficult to replace with a generic AI response:

• Proprietary data: internal statistics, real case studies, sector benchmarks that only your site has
• Authentic and detailed customer reviews: real feedback with photos and precise details cannot be generated by an AI
• Very specialized technical guides: expert content on niche subjects with a depth that general AI cannot achieve
• Comparison with real tests: “we tested these 7 models for 3 months, here are the measured results” — real experience is irreplaceable
• Localized and personalized content: information specific to a market, a city, a precise context

The underlying trend: generic and easily synthesized content is losing its SEO value. Content with an authentic human perspective, real data, and demonstrated expertise gains relative value. This is also what LLMs like ChatGPT and Perplexity prioritize — which is why our GEO approach works on these two surfaces simultaneously.

Stéphane Jambu

SEO engineer, 650+ clients, 1,200+ semantic cocoons deployed from Cambodia. Founder of Hi-Commerce, specialized in e-commerce SEO architecture, GEO (Generative Engine Optimization) and neuroscience applied to content.