Does allowing Googlebot let Google train Gemini on my site?

Googlebot and Google-Extended do different jobs, and Google documents them separately.

• Googlebot decides whether Google can crawl your pages for Search. That covers classic results and Google's AI features in Search, such as AI Overviews and AI Mode.
• Google-Extended decides whether Google can use your content to train and ground Gemini Apps and Vertex AI. Google's common crawlers documentation (updated July 2026) says it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal."

So allowing Googlebot does not, by itself, grant or refuse Gemini training. The absence of a Google-Extended rule does. If your robots.txt allows Googlebot and says nothing about Google-Extended, you have not opted out.

The Google-Extended definition covers the token in more detail, and the Google-Extended vs Googlebot comparison lays out each surface side by side.

As of October 2026, the opt-out is one robots.txt group. Keep your existing Googlebot rules and add:

User-agent: Google-Extended
Disallow: /

That tells Google not to use your content for Gemini Apps and Vertex AI training and grounding. Your Search eligibility stays governed by your Googlebot rules.

Three checks after you deploy it:

1. Confirm the file is live at the root of each host you care about. robots.txt rules apply per host.
2. Confirm you did not accidentally add a broad User-agent: * disallow that also catches Googlebot.
3. Confirm your firewall or CDN is not blocking Google's crawlers outright, which removes you from Search regardless of robots.txt. The Gemini crawler controls page covers the firewall side.

The robots.txt Checker for AI Crawlers at 99Visibility's free tool shows how your file resolves for each named agent. The full pattern for every vendor is in block AI training but stay in AI search.

Google-Extended is not an AI Overviews switch. Google's AI features documentation says a page must be "indexed and eligible to be shown in Google Search with a snippet" to appear as a supporting link in AI Overviews or AI Mode, with "no additional technical requirements." That is Googlebot territory. Google also says robots.txt directives for Googlebot are the control because AI is built into Search.

That creates a trade-off you cannot avoid with robots.txt:

• Blocking Google-Extended keeps you in AI Overviews and AI Mode. See whether blocking Google-Extended removes you from AI Overviews.
• Blocking Googlebot removes you from Search, including those AI features.
• No token sits in between.

If you want to limit how much of a page appears in Google's AI features, the documented on-page controls are noindex, nosnippet, max-snippet and data-nosnippet. These suppress display. They do not make content more likely to be used.

Google also offers a generative AI inclusion control in Search Console. It is separate from both robots.txt tokens.

What the evidence shows about it, as of October 2026:

• It is opt-out. Inclusion is "the default control for all properties," so no action is needed to be eligible.
• It covers AI Overviews, AI Mode and generative AI features in Google Discover.
• Child properties default to "Inherit control from parent," so a property can be excluded without anyone editing it directly.
• Google says it is not a ranking or inclusion signal for the rest of Search.
• It is separate from training. Google-Extended remains the training lever.
• Opting out of display does not stop Google using content to power Search as a whole, for example to understand the language of queries.

That last point matters for anyone who wants a full exit. Read whether Google can still use your content after you opt out before you choose.

A few limits apply whichever setting you choose.

Past training is not undone. A model's parametric knowledge is frozen at its training cutoff. A new Google-Extended rule affects future use, and any effect lands in a later model generation.

The visibility effect is unmeasured. Nothing in the published evidence tests whether opting out of training changes how often a model mentions a brand from memory. The honest label is unmeasured, not harmful and not harmless. If brand answers are your concern, see whether blocking Google-Extended changes what Gemini says about your brand.

Gemini API grounding is a separate path. Grounding with Google Search in the Gemini API is a developer-invoked, separately billed retrieval path. Google documents it apart from Gemini Apps, so check its documentation rather than assuming Google-Extended settings carry over.

Robots.txt states a preference. It is the documented control for compliant crawlers. It is not a technical block.

The next move: decide whether Gemini training matters to you, set Google-Extended to match, and leave Googlebot allowed unless you want out of Search entirely.

Does blocking Google-Extended hurt my Google rankings? Google says no. Its common crawlers documentation states that Google-Extended "does not impact a site's inclusion in Google Search nor is it used as a ranking signal." Your Search eligibility, including AI Overviews and AI Mode, follows your Googlebot rules and normal indexing requirements. Treat this as Google's stated position: it is a first-party assertion, and outsiders cannot independently audit how Google applies it.

Is there a robots.txt rule that keeps me out of AI Overviews but in Search? No. As of October 2026, Google has no separate robots.txt token for AI Overviews or AI Mode. Googlebot is the control, and blocking it removes you from Search entirely. Your options for limiting AI feature display are the snippet controls (nosnippet, max-snippet, data-nosnippet), noindex, and the Search Console generative AI control where it is available to your property.

If I never edit robots.txt, is my site used for Gemini training? If your robots.txt does not disallow Google-Extended, you have not opted out of Gemini Apps and Vertex AI training and grounding. Default inclusion is the starting state. Check every host you run, because robots.txt applies per host, and confirm no template or plugin has added rules you did not intend.

Does allowing Google-Extended make Gemini cite my site more? There is no evidence for that. Google-Extended is a permission, not a ranking or citation signal. Nothing published measures whether allowing training changes how often Gemini names a brand. What you control is access: whether Google can use your content at all. What the model says with that content is outside your control.