For the complete documentation index, see llms.txt. Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.
Skip to content
Attensira LogoAttensira
Book a demo

Bots

Google-GeminiNotebook: allow it or block it?

Google-GeminiNotebook is the fetcher Google uses when a Gemini Notebook user adds your URL as a source: it requests that one page, nothing more. Allow it; each fetch means a person chose your page for their project. It was called Google-NotebookLM, so update old log filters. It generally ignores robots.txt, and it is not Gemini's training control.

Start freeBook a demo

7 days free · card required

AI traffic · attensira.com30 days to 4 Oct 2026

0requests filed as Google-GeminiNotebook · 30 days

of 143,601 AI crawler requests from 7 operators

We do not name Google-GeminiNotebook. It falls through to the wildcard group, which allows everything except /api/. Its operator states robots.txt may not apply to it, so this group is a preference at most.

# No group names# Google-GeminiNotebook, so it# reads the wildcard group.User-agent: *Allow: /Disallow: /api/
Analyst agentAnalyst · reads the AI traffic boardAI traffic board, read over the Attensira MCP, and our live robots.txt · · n=143,601 AI requests
Karl-Gustav KallasmaaKarl-Gustav Kallasmaa, Founder & CEOLast updated

Who runs Google-GeminiNotebook, and what is it for?

Google runs it, and we file it under user-triggered fetch. Reads a page because a person added it as a source in an assistant. Retrieval for that one person, not a crawl and not training.

From Google’s documentation, read . Why a training fetch and a search fetch are different decisions: the AI crawler guide.

How do I allow or block Google-GeminiNotebook in robots.txt?

Paste this into the robots.txt at the root of your domain. The operator says the agent may not read robots.txt, so the group records your preference; enforce anything sensitive on the server.

# Google documents user-triggered# fetchers as generally ignoring# robots.txt. Write this if it# states your intent; do not# expect it to stop the fetches.User-agent: Google-GeminiNotebookDisallow: / # The training control is a# separate token that sends no# requests at all.User-agent: Google-ExtendedDisallow: /

Why does a Google-GeminiNotebook visit look so different from a crawl?

Because it fetches one page for one person. Google documents the Gemini Notebook fetcher as requesting individual URLs that Gemini Notebook users have provided as sources for their projects1. Someone pastes a link into a notebook; this agent retrieves that page. There is no crawl, no link discovery and no sweep of your site.

In a log, the shape is unmistakable: repeated requests on a small number of URLs and nothing on the rest of the domain. If one page is fetched over and over, it is probably a page people keep adding to their notebooks. That is a signal worth having in its own row: someone thought your page was worth working from.

How do I verify that a visit really came from Google-GeminiNotebook?

A user-agent header is free text, so a log line naming Google-GeminiNotebook is a claim until you check it.

Look up the visiting IP, confirm the hostname ends in one of the suffixes in the record, then look that hostname up and confirm it returns the same IP, or match the IP against the published ranges.

Google documents two reverse DNS forms for user-triggered fetchers, *.gae.googleusercontent.com and google-proxy-*.google.com, and splits their ranges between Google-controlled and third-party files. Prefer the range file. Verifying matters more for a token documented as ignoring robots.txt, because an impersonator borrowing the name inherits that claim.

Did the NotebookLM rename break your robots.txt and reports?

Very possibly. Google lists Google-NotebookLM as the former user agent for this fetcher and Google-GeminiNotebook as the current one2. A rename is the quietest failure in crawler measurement. A filter matching the old token does not error and does not warn. It returns zero, and zero looks exactly like "nobody does this any more".

The robots.txt side is the larger effect. Our own robots.txt run across the top 10,000 Tranco domains on 3 September 2026 found 306 of 8,036 readable files blocking Google-NotebookLM at the root, and 19 explicitly allowing it. Those groups name the old token. The fetcher now arrives as Google-GeminiNotebook, which those rules do not mention, so whatever preference those sites wrote no longer names the agent. Because user-triggered fetchers generally ignore robots.txt anyway, the practical cost is in the record of intent rather than in traffic, but it is a record nobody is maintaining.

What to do: match both tokens in logs, keep the old one for comparability, and if you have a NotebookLM group in robots.txt, add the new name beside it.

Does Google-GeminiNotebook follow robots.txt?

Google states that because the fetch was requested by a user, its user-triggered fetchers generally ignore robots.txt rules3. The block in the robots.txt section above says so in a comment: it is a statement of preference, written where preferences belong.

If the fetches really must stop, the control is at the network or application layer, and that is a decision to deny a class of visitor rather than a line in a text file. For most sites it is the wrong decision. A person who adds your page as a source is reading you closely.

VerificationGoogle
Reverse DNS ends in
*.googleusercontent.com*.google.com
IP range file
user-triggered-fetchers-google.json
Researcher agentResearcherTranscribed from Google documentation ·

What user-agent string does Google-GeminiNotebook send?

Match the token, never the version. A rule pinned to a version stops matching silently the next time the operator ships.

The version in this string changes; check Google’s page before relying on the exact text.

Is this how Gemini trains on your content?

No. In the four-way job taxonomy of the AI crawler guide, this is a user-triggered fetch: retrieval for a specific person's project. The documented control over whether Google may use crawled content for training future Gemini models is Google-Extended5, a token that sends no requests and so never appears in a log.

Both have "Gemini" somewhere behind them, which makes the confusion easy, and the result is a rule that answers a question nobody asked. If your legal team wants your content out of Gemini training, the Google-Extended group in the block above is the rule that does it.

Nor is it a citation. Google's documentation describes the fetch and not an output that links back to you, so this page does not claim one.

How is it different from Google's other user-triggered agents?

Google-Agent navigates and acts on a user's behalf. Google-Read-Aloud fetches a page so it can be read out to a user. Google-GeminiNotebook fetches a page a user added as a source. All three are user-triggered, and none is a crawl.

What can Attensira show you about Google-GeminiNotebook?

AI traffic, on Growth and Business, counts which pages AI agents fetch, grouped by operator and declared purpose, measured at the edge, so a page that people keep adding to notebooks stands out from crawling. It reports user agents as sent. How Gemini itself answers about your brand is a separate daily reading, available on Growth and above. See crawler logs and the robots.txt generator.

User-AgentGoogle
Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36 (compatible; Google-GeminiNotebook; +https://developers.google.com/crawling/docs/crawlers-fetchers/google-gemininotebook)
Researcher agentResearcherTranscribed from Google documentation ·

Should I allow Google-GeminiNotebook in my situation?

Concrete cases, and what each one implies for this agent.

  • Your log filters were written against Google-NotebookLM and the token has gone quiet.

    Add Google-GeminiNotebook and keep the old token matching alongside it.

    Google lists Google-NotebookLM as the former user agent for this fetcher and Google-GeminiNotebook as the current one. A rename produces no error anywhere — the filter keeps running and simply reports zero — so the failure looks like a decline in interest rather than a broken rule, which is the most expensive shape a measurement bug can take.

  • A single page is being fetched repeatedly by Google-GeminiNotebook and you want to know why.

    Read it as people adding that URL as a source, not as a crawl of your site.

    Google documents this fetcher as requesting individual URLs that users have provided as sources for their projects. Repeated hits on one URL and none on the rest of the site is the expected shape, and it identifies a page people are actively working with — closer to a bookmark signal than to crawler behaviour.

  • Legal asks you to stop Gemini using your content and someone proposes a Google-GeminiNotebook Disallow.

    Write Google-Extended for the training question. Treat the Gemini Notebook rule as a separate, weaker statement of preference.

    Google documents user-triggered fetchers as generally ignoring robots.txt, so the Gemini Notebook rule may not stop anything, and the fetch it governs is retrieval for a person's project rather than model training. Google-Extended is the documented control over training use, and it is the one a training brief actually needs.

What else do people ask?

What happened to Google-NotebookLM?
Google lists Google-NotebookLM as the former user agent for this fetcher and Google-GeminiNotebook as the current one. Filters and robots.txt groups written for the old name do not match the new one, so update both and keep the old token in reports for history.
Does Google-GeminiNotebook obey robots.txt?
Google states that because the fetch was requested by a user, its user-triggered fetchers generally ignore robots.txt rules. A Disallow states your preference but is not expected to stop the fetch. Stopping it takes a network or application control.
Does blocking Google-GeminiNotebook stop Gemini using my content?
No. The documented control over whether Google may use crawled content to train future Gemini models is Google-Extended, which sends no requests. A Google-GeminiNotebook rule answers a different question and leaves training untouched.
How many sites block NotebookLM?
In Attensira's robots.txt study of the top 10,000 Tranco domains on 3 September 2026, 306 of the 8,036 readable files blocked the old Google-NotebookLM token at the root, about 3.8%. Those groups do not name Google-GeminiNotebook, so they no longer address the fetcher under its current name.

Where do these facts come from?

5 claims on this page, each with the page it was read from; last checked . The numbers in the text point here.

  1. Google documents the Gemini Notebook fetcher as requesting individual URLs that Gemini Notebook users have provided as sources for their projects.

    developers.google.com · retrieved

    “The Gemini Notebook fetcher requests individual URLs that Gemini Notebook users have provided as sources for their projects.”

  2. Google documents Google-GeminiNotebook in mobile and desktop forms whose compatible clause names Google-GeminiNotebook and links to Google's Google-GeminiNotebook documentation page, and lists Google-NotebookLM as the former user agent for the same fetcher.

    developers.google.com · retrieved · changes often, check the source

  3. Google states that because the fetch was requested by a user, its user-triggered fetchers generally ignore robots.txt rules.

    developers.google.com · retrieved

    “Because the fetch was requested by a user, these fetchers generally ignore robots.txt rules.”

  4. Google documents the reverse-DNS forms ***-***-***-***.gae.googleusercontent.com and google-proxy-***-***-***-***.google.com for user-triggered fetchers, and publishes user-triggered-fetchers.json and user-triggered-fetchers-google.json as their address ranges.

    developers.google.com · retrieved

  5. Google documents Google-Extended as the standalone product token publishers use to manage whether content Google crawls may be used for training future generations of Gemini models.

    developers.google.com · retrieved

Want whatever it takes to win the answer?

Start freeBook a demo

7 days free · card required