In short
- AI Overviews and AI Mode use the normal Google index. If a page can appear with a snippet there, it can in principle be used as a source.
- Google says explicitly: there are no additional technical requirements, no special structured data and no AI text files.
- AI Mode breaks questions down into many sub-searches. Pages that answer a sub-question precisely therefore have more chances.
- Traffic from both features is shown in Search Console in the Performance report under “Web”, not separately.
What AI Overviews and AI Mode are
AI Overviews are short AI summaries above the normal search results, with links to the pages the answer is based on. AI Mode is a separate area of Search where users ask follow-up questions and get longer answers, also with sources. Google introduced both features in most European countries in 2024 and 2025, and France followed in July 2026.
For businesses this changes the way people search. Instead of “accountant Leeds”, people type whole questions: “Which accountant in Leeds knows about solar panels on rental properties?”. The AI answers with a few names and links. If you are missing there, you get clicked less often, even if you rank high further down.
How Google chooses the sources
Google describes the process like this: both features can use what is called query fan-out. The question is split into many related searches on subtopics, and the answer is built from the results. This allows Google to link a wider range of pages than in classic search. AI Overviews and AI Mode can use different models and techniques, so the linked pages often differ.
The consequence: it is not only the one page for the main term that counts. A page that clearly answers a sub-question, such as “Roughly what does a heat pump cost in an older house?” or “Which documents do I need for the first consultation?”, can be drawn on as a source for a bigger question.
What Google officially says
Google has published its own guidance for website owners. The key points:
- A page must be indexed and eligible to appear in Search with a snippet. There are no further technical requirements.
- No new machine-readable files, no AI text files and no special Schema.org data are needed.
- The familiar basics apply: allow crawling, internal links, important content as text, a good page experience, structured data that matches the visible content.
- If you want to show less, use
nosnippet,data-nosnippet,max-snippetornoindex. These settings apply to Search as a whole.
There is no guarantee of being included, not even with a perfect implementation.
Six levers for businesses
- Check indexing: service pages must be in Google’s index. The URL Inspection tool in Search Console shows whether Google knows the page and how it sees it.
- Answer sub-questions: add sections on cost, process, duration, requirements and service area to your service pages, in full sentences and with concrete details.
- Answer first: start sections with the core of the answer and explain afterwards. That way a paragraph works well as evidence.
- Show evidence: give figures, standards, certificates and sources. Statements backed by evidence come across as more reliable, to people and machines alike.
- Maintain local signals: for questions with a location, the Google Business Profile, reviews and consistent company data count. More in our Business Profile checklist.
- Keep structured data clean: it is not a requirement, but it helps describe organisation, location and services unambiguously. What matters is that it matches the visible text. More in the article on structured data.
Measuring: what Search Console shows
Search Console counts clicks and impressions from AI Overviews and AI Mode as normal search traffic, in the Performance report under the search type “Web”. There is no separate column for them. It is therefore worth looking at two things: pages whose impressions rise while clicks fall, and new queries phrased as questions.
Search Console does not show whether your business is named in the answers. For that you have to look at the answers yourself, several times and for several questions, because they vary. That is exactly what our free Quick Check automates for ChatGPT and Gemini, and the Location Analysis for five systems including Google’s AI Overviews.
What does not work
- Text for the AI instead of for people: paragraphs stuffed with keywords or hidden instructions to AI systems violate Google’s spam policies.
- Mass-producing pages with AI: lots of thin pages with no value of their own tend to do harm. Google assesses usefulness, not how content was created.
- Relying on a single measurement: one answer today says little about tomorrow. What counts is the share across many questions and runs.
When help pays off
You can tackle the basics yourself. Help pays off when you want to know which questions are asked in your industry, which sources Google and other AI systems draw on and where your website has gaps. Our Location Analysis asks five AI systems, including Google (AI Mode and AI Overviews), 20 customer questions three times each and delivers the ten most important measures, for €229 plus VAT. For several countries or locations there is the Market Analysis from €890 plus VAT. More on our GEO agency page.
Sources
- Google Search Central: AI features and your website
- Google: AI Mode in Google Search expands to more languages and locations
- Blog du Modérateur: Google déploie AI Overviews et AI Mode en France
- Google Search Central: Spam policies for Google web search
- Google Search Central: Creating helpful, reliable, people-first content