Short answer: AI Overviews draw on Google's ordinary index, so there is no separate crawler to allow. The one lever Google documents is Google-Extended, and it is not a crawler at all — it is a usage control.
Google-Extended is the cleanest separation any vendor offers
Google documents Google-Extended as controlling whether content it has already crawled may be used to train Gemini models and to ground answers in Gemini Apps and Vertex AI — and states explicitly that it does not affect inclusion or ranking in Google Search. That makes it the only place in this whole category where the trade-off can be reasoned about instead of guessed at.
Two consequences people miss:
- You cannot verify it in server logs. It governs usage, not fetching, so nothing shows up as a request. Treat it as a stated policy decision, record it, and measure the effect on the answer side.
- Disallowing it has a bounded cost — grounded Gemini answers — not the vague sitewide cost people assume.
What actually earns a place in the Overview
Because the index is shared, the ordinary work still applies: the page has to be crawlable, server-rendered, and indexed. Beyond that, the Overview is assembled from passages, which puts the emphasis on extraction:
- One question per heading, phrased the way a person would ask it.
- The answer in the first sentence under that heading, with the subject named — not a pronoun that depends on the heading.
- Facts in a table with real units and explicit dates. "Recently" and "currently" cannot be quoted.
- FAQ content in the HTML, not only in the schema. Schema over hidden content is the most common self-inflicted wound here.
Measure it as presence, not position
There is no position ten in an Overview. There is named, mentioned in passing, or absent — and separately, cited or not cited.
Overviews also vary between runs and between accounts, so a single check proves nothing. Use a fixed question set, record which runs produced an Overview at all, and never render a run that failed as a zero.