What decides whether you appear in Google's AI Overviews

AI Overviews changed the geometry of a Google results page. The summary sits above everything, answers a good share of questions on the spot, and the pages it links become the shortlist a searcher actually sees. Which raises the operational question: what gets a page into that shortlist?

Google has answered more of this than people assume — in its ordinary search documentation, not in a leak or a patent reading.

The documented mechanism

Google's guidance on AI features states it directly: "There are no additional requirements to appear in AI Overviews or AI Mode." No registration, no new file format, no AI-specific markup. A page qualifies as a potential citation if it is indexed and eligible to appear in Search with a snippet — the same bar ordinary results clear.

That sentence quietly invalidates a lot of what is currently sold as "AI Overview optimization." If a pitch involves a special machine-readable file or proprietary markup as the price of entry, the pitch contradicts the platform's own documentation.

Eligibility is not selection, though. Google has described AI Mode as running a query fan-out — issuing multiple related searches behind one question and drawing sources from across them. A page that squarely answers one of those sub-questions can be cited on a query it never ranked first for. The competitive surface is wider than the ten blue links were: a page competes on every sub-question it can answer, not just the head term.

The controls that actually apply

What AI features may show from a page is governed by the snippet controls that have existed for years: nosnippet, data-nosnippet, max-snippet, and noindex. Those are the levers Google documents for limiting how content appears in AI Overviews — the same levers that govern ordinary snippets. A max-snippet limit someone set years ago for a licensing reason now also caps what AI features can quote from you.

The Google-Extended mistake

Here is where teams that did their homework still get burned. Google publishes a control called Google-Extended, and blocking it looks like the obvious way to opt out of "Google's AI." It is not. Google-Extended governs whether content feeds training and grounding in Gemini-side systems. AI Overviews eligibility follows Search indexing and snippet rules — a different pipeline. The misreading runs in both directions: sites block Google-Extended expecting to vanish from AI Overviews (they do not), and sites hesitate to block it fearing their search traffic depends on it (it does not).

What to do with this

• Treat AI Overview presence as a function of classic eligibility first: indexed, snippet-allowed, ranking within reach for the query set.
• Audit snippet controls deliberately — old nosnippet and max-snippet decisions now have AI-feature consequences their authors never intended.
• Decide the Google-Extended question on its own merits. It is a training-policy and IP decision, not a visibility lever.
• Measure with care: Search Console folds AI Overview impressions into ordinary Search totals, so shifts in position and CTR need reading against feature presence, not in isolation.

The mechanism is classic search wearing a new interface.