Google AI Overviews
AI Overviews sit on top of Google Search, draw on the same index, and require no new markup of any kind. Google's own documentation is unusually blunt about the last point.
What is Google AI Overviews?
Google AI Overviews is a composed answer that appears within Google Search results, above or among the ordinary ranked links, with source links attached. Unlike the other systems documented here, it is not a separate destination. A reader does not go to it; it arrives in the middle of a search they were already doing.
That placement is the most consequential fact about it. The other three engines compete for a visit. AI Overviews changes the shape of a page the reader was already on, which means its effect on publisher traffic is felt without anyone changing their habits.
How does Google select the pages it shows?
Google documents one substantive mechanism, which it calls query fan-out. Rather than running the user's question as a single query, the system issues multiple related searches across subtopics of that question, and draws candidate pages from across those searches. Google's stated reason for this design is that it surfaces a wider and more diverse set of links than a single query would return.
The practical reading for a publisher is that relevance to the literal question matters less than it would on a results page. A page that comprehensively covers the neighborhood around a question can be pulled in through a subtopic search even if it would not have ranked for the question as typed.
Beyond fan-out, Google does not publish how a retrieved page becomes a cited one. There is no documented scoring, no stated threshold, no explanation of why one of two similar pages is named. Anyone describing that process with confidence is inferring.
What markup does a page need to appear?
None, and Google says so directly. Its documentation states that site owners do not need to create new machine readable files, AI text files or markup to appear in these features, and that there is no special schema.org structured data to add. Pages need only meet the ordinary requirements for being indexed in Google Search.
This single paragraph of official documentation contradicts a substantial amount of paid advice, including various proposed standards files and AI specific markup schemes sold as prerequisites. Google's stated path to appearing in its AI features is foundational SEO.
What control does a site owner have?
Control runs through the preview controls publishers already use, not through
anything new. The nosnippet, data-nosnippet and
max-snippet directives, along with noindex, govern how
content can appear in AI features in the same way they govern ordinary snippets. A
publisher who restricts snippet display is also restricting AI feature inclusion,
because the same directives do both jobs.
Separately, the Google-Extended control limits the use of a site's
content for training other Google systems. It is worth being precise about the
difference: preview controls affect whether your content appears in an AI Overview,
while Google-Extended concerns training. They are not interchangeable, and blocking
one does not accomplish the other.
What is not publicly known?
Several things that matter.
Google does not publish which queries trigger an AI Overview, so a publisher cannot determine in advance whether a topic is affected. It does not publish how many sources a given overview will cite, or how that number is decided. It does not publish how a source is chosen for citation once retrieved, beyond the fan-out description of how candidates are gathered. And it does not publish how often the feature is shown, which means industry figures on coverage are measurements by third parties with their own sampling methods, not reported numbers.
It is also worth noting that this documentation carries a last updated date and the behavior it describes has changed repeatedly since AI Overviews launched. Anything you read about this feature, including this page, should be checked against the current version of Google's own documentation before you act on it.
How does this interact with ordinary ranking?
It sits on top of it rather than replacing it, and the practical consequence is that the two cannot be worked on separately.
Google's documentation is explicit that pages need only meet ordinary Search indexing requirements to be eligible for AI features, and that foundational SEO is the pathway to visibility in them. There is no second pipeline with its own entry requirements. A page that cannot be indexed cannot be cited, and a page that is invisible for the subtopics around a question will not be gathered by a fan-out that searches those subtopics.
What differs is which pages benefit. Fan-out gathers candidates across several related searches, so breadth of coverage around a topic matters more than it would for a single competitive term. A site with thorough, separately addressable coverage of the questions adjacent to a topic has more surfaces to be gathered from than a site with one heavily optimized page aimed at the main query.
The preview controls add a tension worth naming. Restricting snippets protects content from being shown without a click, and it also restricts appearance in AI features, because the same directives govern both. A publisher cannot currently allow rich AI feature inclusion while suppressing ordinary snippets, or the reverse. That is a single decision with two consequences, and it should be made as one.