Google documents how robots.txt and meta tags decide exposure to AI Overviews and AI Mode

Official Google Search Central guide, updated in December 2025, brings together in one place the technical directives that control whether a page enters the responses generated by AI Overviews and AI Mode.

Editor's note: this article states that the guide was updated on December 31, 2025. The official source records the update date as December 10, 2025. We kept the rest of the text, which was verified and matches Google's documentation, but readers should consider December 10 as the correct update date for the guide.

Google Search Central published a consolidated guide called "AI features and your website," updated on December 31, 2025, bringing together in a single document the technical rules that determine whether a page can appear as a supporting link in AI Overviews and AI Mode. It isn't a new feature: it's the formalization, in official documentation language, of something that until then was scattered across blog posts and one-off answers from Google itself.

For those working with GEO (generative engine optimization) in Brazil, this changes little about the strategy, but changes a lot about how to justify technical decisions to the team and to the client: now there is a single, citable source from Google explaining exactly which directives control presence in AI responses.

The technical requirements are the same as for traditional search

The guide's most direct point is also the most important one: there is no additional technical requirement to appear in AI Overviews or AI Mode beyond already appearing in regular search. Google's rule is simple: the page needs to be indexed and eligible to display a snippet in traditional Search, meeting the standard technical requirements (crawlable, not blocked by robots.txt, no indexing errors).

This debunks a common myth in GEO discussions: there is no parallel "AI mode" with its own schema rules, exclusive meta tag, or special signal that unlocks inclusion. In short: if a page doesn't appear in regular organic search with a snippet, it also won't appear in AI Overviews or AI Mode, because the pool of candidates is the same index.

The guide also reinforces the usual best practices: allowing crawling in robots.txt (including CDN and hosting infrastructure), having internal links that make content easy to find, offering a good page experience, keeping relevant textual content available in plain text, using quality images and videos when it makes sense, and ensuring structured data matches the text visible on the page.

The directives that decide exposure

The practical core of the document lies in the "preview" controls, the directives that limit what is shown from a page that's already indexed. These are the same directives, already familiar to those who do traditional SEO, that Google now explicitly confirms as valid for blocking exposure in AI responses.

A point that causes confusion and that the guide clarifies well: Google-Extended is a user-agent separate from Googlebot, used to control training and grounding for other Google AI systems, not presence in AI Overviews and AI Mode within Search. Blocking Google-Extended in robots.txt doesn't remove the page from Search's AI responses; whoever wants that needs to act on Googlebot with noindex or nosnippet. It's a common mistake to confuse the two and think you've blocked exposure when you've only blocked training for another product.

How the response is assembled under the hood

The document describes that AI Overviews and AI Mode can use a technique called "query fan-out": the system fires off multiple related searches, covering different subtopics and data sources, to assemble a more complete response. While the response is being generated, Google's models identify additional supporting pages, which explains why the set of links cited in an AI response tends to be broader and more diverse than that of a classic search.

This has a direct practical implication for writers: a page can be cited in an AI response not for the main keyword it targets, but for answering well a side subtopic that the fan-out decided to explore. Content that covers multiple angles of a subject, instead of optimizing only for a central query, has a better chance of being picked up in one of these parallel firings.

It's worth noting that AI Overviews only appears when Google's system determines that it adds something beyond classic search, which is why it "often isn't triggered," in the guide's own words. There is no guaranteed trigger by content category or query type.

What to measure before and after the change

Here's the point most relevant to those who need to prove impact. Sites that appear in AI Overviews and AI Mode are counted within overall search traffic in Search Console, specifically in the Performance report, under the "Web" search type. As of this documentation, there is no separate filter for "clicks coming from AI Overviews" within Search Console.

This means measuring GEO impact requires an indirect process:

1. Establish a baseline of impressions and clicks in the Web report before any change to preview directives.
2. Apply the change (for example, removing an unnecessary nosnippet) and wait for the recrawl, which can take anywhere from days to months.
3. Track the evolution of impressions per page, since AI Overviews tends to generate more impressions without a proportional click rate (the user already got the answer right on the SERP).
4. Cross-reference with Google Analytics (or another tool), looking at time on site and conversion, not just click volume.

The guide brings a relevant qualitative data point here: according to Google, clicks originating from results pages with AI Overviews tend to be of higher quality, in the sense that the user usually spends more time on the site after clicking.

For practical dashboard purposes, this suggests tracking engagement rate and session time as a complementary metric to traditional CTR, since CTR alone tends to drop when the response satisfies a good part of the user's intent right on the results page.

When the directive doesn't work: a common misstep

The guide lists a troubleshooting roadmap for when a preview directive (such as nosnippet or noindex) has been applied and the content still appears in AI features:

• Confirm that the directive is correct and visible to Googlebot, using Search Console's URL Inspection tool to see the HTML that Googlebot actually received.
• Allow time for the recrawl: Google warns this can take anywhere from a few days to several months, depending on how often that page is crawled.
• Request manual recrawling when the change is urgent.
• Take the case to the Google Search Central help community if none of this resolves it.

A recurring mistake in these cases is the persistence of an outdated version of the HTML without the directive applied, something that only the URL Inspection test reveals clearly.

What remains open

The document is clear about what controls exposure, but doesn't detail ranking criteria within AI responses, nor how the fan-out weighs one source against another. For those doing GEO in Brazil, the practical consequence is that there is no AI-specific optimization checklist beyond doing solid technical SEO and ensuring that preview directives reflect the site's actual intent: whoever wants to appear, opens access; whoever wants to limit it, uses nosnippet, max-snippet, or noindex, knowing that the same directive applies to all of Search, not just AI.