Google says AI Overviews does not require special structured markup

Google Search Central's documentation, updated in December 2025, is direct: there is no additional technical requirement for appearing in AI Overviews and AI Mode responses. What changes is how to measure and control that presence.

Editor's note: the text originally cited the documentation's update date as December 31, 2025; the official source records 'Last updated 2025-12-10 UTC' (December 10). We kept the rest of the content because the central technical claims (absence of additional requirements for AI Overviews, eligibility criteria, query fan-out, crawling controls) are faithfully supported by the documentation, but the specific date does not hold up. In addition, the individual description of each control's effects in the table below (nosnippet, data-nosnippet, max-snippet, noindex) is our own elaboration on the known behavior of these directives, and is not detailed on the source page consulted, which only lists the four names.

For technical SEO professionals following the growth of AI Overviews in Portuguese-language queries, one question has taken over technical SEO meetings: is there any specific structured markup, some new schema, that gets a page pulled into the AI-generated response? Google Search Central's documentation on the topic, updated on December 31, 2025, answers directly: there isn't one. The technical criteria for appearing as a supporting link in AI Overviews or AI Mode are exactly the same ones already used for classic search.

What the documentation actually requires

Search Central's hub page on AI features sums up the technical requirement in a sentence worth quoting in full, because it closes the door on any theory of exclusive schema markup:

To appear as a supporting link in AI Overviews or AI Mode, a page needs to be indexed and eligible to appear in Google Search with a snippet, meeting Search's technical requirements. There are no additional technical requirements.

Official Google Search Central documentation

In practice, this means that noindex, robots.txt blocks, or lack of snippet eligibility remain the only real technical barriers. There is no schema.org attribute that unlocks presence in the generated response, nor a new tag to "activate" the feature.

The practices that still apply (and aren't new)

The documentation is explicit in saying that no specific optimization is needed for AI Overviews and AI Mode: everything that was already a foundation of SEO continues to be what matters. Google lists points that were already in the playbook of any technical audit:

• Making sure crawling is allowed in robots.txt and by any CDN or hosting infrastructure
• Making content easy to find via internal links
• Providing a good page experience for users
• Making sure important content is available in text format
• Supporting text content with quality images and videos, when applicable
• Making sure structured data matches the page's visible text
• Keeping Merchant Center and Business Profile information up to date

None of these items is exclusive to generative AI. They are the same pillars that already appeared in traditional SEO guidelines, just now gathered in one place as a checklist for those who also want to appear in synthesized responses.

Structured data: consistency, not an extra checklist

The most relevant point for those who work with structured data is subtle and easy to miss among the other items on the list: the requirement isn't to have more schema, it's to have schema that matches what's on screen. Imagine, for illustration purposes, a product marked up with Product and offers.price at R$ 199, but displayed on the page at R$ 249 after a promotion that didn't update the JSON-LD: that's the kind of hypothetical inconsistency that compromises eligibility, long before any discussion about AI.

This reinforces a practice that every technical SEO audit should automate: validating structured data against the rendered DOM, not against the static template. Tools like Google's own structured data testing tool and programmatic checks in CI remain the way forward, not a new type of markup designed for AI.

Core Web Vitals: what the source says, and what it doesn't detail

It's worth being honest about the documentation's limits here. The text cites "providing a great page experience for users" as a general best practice, but does not specify which Core Web Vitals metrics (LCP, INP, CLS) directly weigh into page selection for AI Overviews, nor does it publish its own thresholds for this feature. Anyone expecting a table of thresholds exclusive to generative AI won't find one in this documentation.

What can safely be inferred: since eligibility depends on the page already being qualified to appear with a snippet in classic search, the same page-experience signals that affect that snippet remain relevant. There is not, as of this documentation update, a Core Web Vitals metric treated as a specific gate for AI Overviews.

How the response is assembled under the hood

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 put together a response. While the response is generated, Google's models identify additional supporting pages, which allows for displaying a broader and more diverse set of links than in a classic search.

This explains why a page can appear as a source in an AI response without ever having held the top organic position for the main query: it may have been captured by one of the fan-out's sub-searches, answering a specific slice of the question. For those working in technical SEO, this expands the number of target queries worth monitoring, including long-tail variations that previously seemed irrelevant.

Controls for those who want to get into (or out of) the response

Since AI is embedded in Search itself, robots.txt directives for Googlebot remain the main crawling control. To limit what appears from a specific page, Google lists four mechanisms:

To limit model training and use in grounding for other Google systems, the documentation points to Google-Extended, a control distinct from the four listed above, aimed specifically at AI training and grounding.

Measuring the impact in Search Console

Pages that appear in AI features are counted within overall search traffic in Search Console, reported in the Performance report, under the "Web" search type. There is no report segregated just for AI Overviews in this documentation; the signal is mixed in with the rest of the search data, which requires comparing trends over time rather than looking at an isolated number.

One data point worth noting for those designing the tracking dashboard: Google itself states it has observed that clicks coming from result pages with AI Overviews tend to be higher quality, in the sense that users spend more time on the site after clicking. It's a signal to cross-reference with Google Analytics or another behavior tool, not a number provided directly in Search Console.

Troubleshooting when the control doesn't work

If a page has nosnippet, noindex, or another control implemented and still keeps appearing in AI features, the suggested diagnostic script is:

1. Confirm that the control is correct and visible to Googlebot, using the URL Inspection tool to see the HTML the crawler actually received
2. Give Google time to recrawl the page, a process that can take from days to months depending on the site's update frequency
3. Request a new crawl via Search Console when the change is urgent
4. Bring the case to the Google Search Central help community if the problem persists after the previous steps

What changes in practice for those doing SEO in Brazil

For those handling technical SEO in Brazil, the job isn't to create a new layer of markup: it's to go back to the basics and make sure they're flawless. That means auditing the consistency between structured data and rendered content, confirming that important pages aren't being unintentionally blocked by robots.txt or noindex, and reviewing page experience with the usual metrics.

The measurement angle changes more than the technical one: teams should track the Performance report by segmenting by query type (informational, comparative, long-tail), since query fan-out favors precisely that kind of search. Comparing clicks and time on page before and after relevant changes in AI Overviews coverage is the most honest experiment available, given that Google doesn't yet offer an isolated cut for that traffic.