AI opt-out (Google Search)

What Google says

The sentence is in Google's documentation on AI features, and it is unusually direct:

"AI is built into Search and integral to how Search functions, which is why robots.txt directives for Googlebot is the control for site owners to manage access to how their sites are crawled for Search."

Read that as the answer to the question people actually ask. There is no AI-Overviews: none. The control is the one that has always existed, and it is all or nothing: allow Googlebot and you are in Search with its AI features, or disallow it and you are in neither.

The same page names what you can limit:

"To limit the information shown from your pages in Search, use nosnippet, data-nosnippet, max-snippet, or noindex controls. To limit AI training and grounding in some of Google's other systems, read more about Google-Extended."

Two different things sit in those two sentences, and putting them in one paragraph is what causes the confusion.

What the snippet controls actually reach

The robots documentation is more specific than the AI features page, and it names the AI surfaces outright. On nosnippet:

"Do not show a text snippet or video preview in the search results for this page. […] This applies to all forms of search results (at Google: web search, Google Images, Discover, AI Overviews, AI Mode) and will also prevent the content from being used as a direct input for AI Overviews and AI Mode."

So nosnippet does reach AI Overviews, and not only in what gets displayed: the sentence says the content will not be used as a direct input. max-snippet carries the same clause in a weaker form, limiting "how much of the content may be used as a direct input for AI Overviews and AI Mode".

That is the opposite of how these rules are usually summarised, and it still does not hand you the switch. Read the same sentence for what it costs. The rule applies to all forms of search results, web search included, in one breath with the AI surfaces. Nothing in the wording separates the two, and no second rule does it either.

data-nosnippet narrows the target from the page to a section, which changes the size of the cost without changing the mechanism. noindex removes the page from Search altogether.

So the honest summary is narrower than the usual one: the control that reaches AI takes your search snippet with it, and the documentation offers nothing that separates them.

One exception is written into the same page, and it runs the other way. Structured data is not covered: Google says robots meta tag limitations do not affect the use of schema.org data, apart from description values, and that "structured data remains usable for search results when declared within a data-nosnippet element". A page can carry nosnippet and still feed rich results from its JSON-LD.

The one real opt-out, and where it applies

Google-Extended is the control people are usually looking for, and it does not work where they are looking. From the crawler documentation:

"Google-Extended is a standalone product token that web publishers can use to manage whether content Google crawls from their sites may be used for training future generations of Gemini models that power Gemini Apps and Vertex AI API for Gemini and for grounding (providing content from the Google Search index to the model at prompt time to improve factuality and relevancy) in Gemini Apps and Grounding with Google Search on Vertex AI."

And then the sentence that decides where it applies:

"Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search."

So it covers training of Gemini models and grounding in Gemini Apps and on Vertex AI. It does not cover AI Overviews or AI Mode, which live in Search. Blocking it changes nothing about what appears above the blue links.

You cannot see whether it is honoured

One more line from the same documentation deserves attention, because it removes a verification most site owners would assume they have:

"Google-Extended doesn't have a separate HTTP request user agent string. Crawling is done with existing Google user agent strings; the robots.txt user-agent token is used in a control capacity."

There is no Google-Extended in your logs. The crawl happens under the ordinary Google agents, and the token only tells Google what to do with what it already fetched. You can set the directive, and you cannot measure its effect from your side.

What to do with it

Stop looking for the free switch. You can keep your content out of AI Overviews as a direct input. You cannot do that and keep your search snippet, because one rule does both.

Decide the display question separately from the training question. They use different controls, apply to different products, and answering one does not answer the other.

Price nosnippet before you use it as an AI measure. Take the clicks your snippet earns today and ask whether stopping the AI use is worth that much. For most sites it is not, and that is the whole decision.

Set Google-Extended for what it actually governs. Training of Gemini models and grounding outside Search. If that is your concern, it is the right instrument, and it will not show up in your logs either way.