There is a lot of advice about “optimising for AI Overviews”. Most of it is inference. Google has now written down what it wants, in two developer pages and a help article. This article sticks to those documents and quotes them where it matters.
The one-line answer
Google says the requirements for AI Overviews and AI Mode are the same as for Search: be indexed and be eligible for a snippet. Everything else is ordinary SEO.
The AI features page puts it plainly: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
How do AI Overviews and AI Mode find pages?
Google’s optimisation guide states that “our generative AI features on Google Search are rooted in our core Search ranking and quality systems”. It names two techniques.
The first is retrieval-augmented generation, which the guide also calls grounding. It is “a technique used to improve the quality, accuracy, and freshness of AI responses” by pulling pages from the Search index.
The second is query fan-out. The guide defines it as “a set of concurrent, related queries generated by the model to request more information”. The AI features page adds that this lets Google show “a wider and more diverse set of links” than a classic result page.
Our reading of this is that a page that is retrievable for a narrow sub-question can be cited for a broad one. You do not have to rank for the head term to be a source.
What does Google say you do not need?
No new files, no special markup and no splitting pages into fragments. This is the part the industry resists, so here it is verbatim.
- “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search.” That covers llms.txt and similar proposals.
- “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” Structured data still matters for rich results, which is a separate benefit.
- “There’s no requirement to break your content into tiny pieces for AI to better understand it.” Google says it can understand several topics on one page.
The guide also addresses the labels. It says “optimizing for generative AI search is optimizing for the search experience, and thus still SEO”. We use the phrase generative engine optimization on this site because clients search for it. We do not treat it as a separate discipline, and neither does Google.
What Google says you should do
The guidance is short and recognisable.
Make the content crawlable. The AI features page tells you to confirm Googlebot is allowed in robots.txt. Check that no CDN or host blocks it either. Keep important content as text rather than inside images or scripts.
Write non-commodity content. The optimisation guide defines this as content that “provides unique expert or experienced takes that go beyond common knowledge and the ordinary”. Generic restatements of widely known information are the opposite.
Keep structured data honest. Where you use it, it must match the visible page. This is a general Google policy, and it is the rule behind our refusal to add schema that the page does not support.
Leave snippets on. The controls that limit snippets also limit AI features. See the next section.
The controls, and what they actually do
Four controls appear in Google’s documentation. They are often confused, so here is the difference.
Crawled
Googlebot fetches the page.
Stopped by: robots.txt, CDN or WAF blocks
Indexed
The page is in the Search index.
Stopped by: noindex
Snippet-eligible
Search may show text from it.
Limited by: nosnippet, max-snippet, data-nosnippet
AI Overviews and AI Mode
Shown when Google’s systems decide it helps.
Stopped by: the Search Console generative AI control
Separate: Google-Extended
A robots.txt token for Gemini training and grounding. Google says it does not affect inclusion in Search.
| Control | Where | Effect, per Google |
|---|---|---|
nosnippet, max-snippet, data-nosnippet |
Robots meta tag or HTML attribute | Limit what Search shows from the page, including in AI features |
noindex |
Robots meta tag | Removes the page from Search, and so from AI features |
| Search generative AI control | Search Console setting | Excludes the site from AI Overviews, AI Mode and Discover AI features |
Google-Extended |
robots.txt token | Controls use for Gemini training and grounding in Gemini Apps and Vertex AI |
The Search Console control is the one site owners ask about most. Google says that when a site is excluded, “your site’s content is prevented from being visible to users in Search generative AI features”. It adds that “content crawled from your site won’t be eligible to be used as an input to generate an AI response”. The help page notes that this does not affect training, which is handled by Google-Extended, and does not affect ranking in classic Search.
The common crawlers page confirms the other side. “Google-Extended does not impact a site’s inclusion in Google Search nor is it used as a ranking signal in Google Search.” Blocking it is a training decision, not a visibility decision.
Google also warns that changes to these controls are not instant. The AI features page says “crawling can take anywhere from several days to several months”.
How do you measure it?
With Search Console. It now has a Generative AI performance report covering impressions in AI Overviews and AI Mode. Google says AI feature appearances are also counted inside the “Web” search type of the main Performance report. We cover what those reports do and do not show in our article on measuring AI visibility.
A checklist built only from Google’s pages
Each line below maps to a statement in the sources linked in this article. Nothing here is inferred.
- The page is indexed and eligible for a snippet.
- Googlebot is allowed in robots.txt, and no CDN or WAF rule blocks it.
- Important content is HTML text, not image or script-only.
- No
nosnippetor restrictivemax-snippeton pages you want quoted. - The Search Console generative AI control is set to include the site, if that is your intent.
- Structured data, where present, matches the visible content.
- The content says something an expert would say and a generic page would not.
- You have decided on
Google-Extendedseparately, as a training question. - You are checking the Generative AI performance report, and you understand it reports impressions.
What we saw on our own properties
Across the eight sites in our case studies, the share of Google impressions recorded in Search Console’s Generative AI performance report ranged from 1.4% to 17.0%. The pages shown were rarely only the home page:
- On onlinephotoprinting.co.nz, 928 AI-feature impressions were spread across 553 pages; the home page, the most-shown, had 56.
- On woodcharcoal.in, the three most-shown pages were a category page (712), a comparison page (489) and a product page (468), across 89 pages in all.
- On firebrickssource.com, the standards page led (72), ahead of two product pages.
That spread is what you would expect from the query fan-out Google describes: specific pages answering specific sub-questions. Search Console does not show which query or passage produced each impression, so we report the pattern and not a cause.
What we don’t claim
Meeting every item above makes a page eligible. Google states that AI Overviews are shown only when its systems decide they add to classic results, so eligibility is not appearance. We do not promise placement in AI Overviews or AI Mode. We can make sure nothing on your side is stopping it. We can also show you how to read the reports that tell you what happened.