BackCrawlers

Google-GeminiNotebook: What It Fetches and How to Track It

Google-GeminiNotebook fetches Gemini Notebook sources, generally ignores robots.txt, and can show in Aiola Analytics under Answers.

Panagiotis Visilias

Oct 26, 2026 · 6 min read

Google-GeminiNotebook crawler identification and tracking

Google-GeminiNotebook fetches URLs that Gemini Notebook users add as project sources. Google lists it with user-triggered fetchers. Aiola Analytics can store those hits when your server or edge reports them.

The NotebookLM token was renamed

Google's July 16, 2026 changelog changed the NotebookLM user agent to Google-GeminiNotebook. The same note says NotebookLM became Gemini Notebook.

The July note said Google would keep supporting the old value during the transition. The fetcher listing calls Google-NotebookLM a former agent, supported until August 2026.

If you hardcoded Google-NotebookLM, Google says to update that string. Match Google-GeminiNotebook, not the search spelling google-geminibotnotebook. That extra word is not in Google's token.

Published headers fetch one source URL

Google documents a mobile header and a desktop header. Both contain Google-GeminiNotebook. The fetcher requests individual URLs supplied as project sources.

The desktop example on that page is:

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)

The mobile example uses Chrome/138 and the same token. Match that token inside the full header. The current examples are browser-like user agents. The former token is the short string Google-NotebookLM.

These four Google names do different work

Common crawler documentation defines Googlebot, GoogleOther, and Google-Extended. The fetcher page defines Google-GeminiNotebook. Those pages say their lists are not exhaustive.

IdentityDocumented roleKeep this separate
Google-GeminiNotebookFetches URLs a Gemini Notebook user supplied as sourcesUser-requested source fetch
GooglebotCrawls pages for Google Search indexingSearch indexing
Google-Extendedrobots.txt control for Gemini training and groundingNo separate HTTP user agent
GoogleOtherGeneric crawler product teams may use, including internal researchNot tied to one product

Crawling preferences for Googlebot affect Google Search, including Discover, Images, Video, and News.

Google-Extended has no HTTP user agent of its own. Google crawls with existing user agents. Publishers use the token to manage future Gemini training and grounding. It does not change Search inclusion or ranking.

GoogleOther may fetch public pages for product teams. One example Google gives is a one-off crawl for internal research. Preferences for GoogleOther do not target a named product.

See the token and source-fetch note on Aiola's Google-GeminiNotebook reference.

Robots.txt generally does not stop this fetch

User-triggered fetchers generally ignore robots.txt, because a person requested the fetch. That is the fetcher page's rule. Common crawlers always obey robots.txt when crawling automatically.

A disallow for Google-GeminiNotebook does not guarantee that these source requests stop. Compare the rule you published with the requests your logs still show.

Google-Extended is a different control. It manages training and grounding for content Google already crawls. It is not a request user agent for this fetcher. A Google-Extended line does not replace the user-triggered rule above.

Confirm the address before you trust the name

Google warns that a user-agent string can be spoofed. Its verification guide gives a manual check and a range check.

Use the source IP from your server or edge log.

  1. Run a reverse DNS lookup on that IP.
  2. Confirm the domain is googlebot.com, google.com, or googleusercontent.com.
  3. Run a forward lookup on the hostname from step 1.
  4. Require that lookup to return the original IP.

User-triggered fetchers share published ranges. Google names user-triggered-fetchers.json, user-triggered-fetchers-google.json, and user-triggered-agents.json. Reverse DNS masks on that page use gae.googleusercontent.com or a google-proxy name on google.com.

The Gemini Notebook section does not name a range file of its own. A match on a shared list shows a published user-triggered address. It does not name this fetcher by itself.

Aiola's crawler record for Google-GeminiNotebook has no dedicated IP range. Do the address check in your logs.

Aiola lists both tokens in Answers

The AI crawler reference lists Google-GeminiNotebook, formerly Google-NotebookLM, in Answers. Answers means an assistant fetched a page on demand for a live user. The supported list in Set Up Analytics includes both tokens.

Analytics stores a hit only for a known crawler user agent, and the host must match the site domain. An unknown user agent returns HTTP 204 and is not stored. Aiola derives the bot, vendor, and kind from the user agent. Callers cannot choose that classification.

A fetch is shown in Answers. The AI view shows bots, pages, crawler 404s, and change over time. Aiola Analytics names 70 crawlers. Human visits stay in the Traffic view.

Crawlers do not run JavaScript, so the browser tag cannot see them. Choose one server or edge setup that posts the user agent to the bot ingest endpoint.

Open Settings → Analytics in the project you want to track. If analytics is off, enter the production hostname and select Enable Analytics. Use a bare hostname, with no scheme, path, or port.

The documented payload is a tracking ID, path, host, user agent, optional country, and optional status. There is no source IP field. A stored 200 can mean the capture omitted status, which then defaults to 200.

A row in Answers means an identified assistant fetcher requested that URL. No row means Aiola got no matching beacon in that period. That absence does not prove the fetcher cannot reach the site.

Hosted Analytics needs Pro or the seven-day trial. The trial is for a first-time subscriber with no prior subscription, on Pro Monthly at 100,000 events. Checkout collects a card. Other monthly tiers, Pro Yearly, and GUI Lifetime have no trial. See Plans and Access.

Without Pro or an active trial, browser and bot collectors return 402 and do not store the event.

Desktop version 2.0.8 is the current Aiola release. It runs on macOS 12+ and Windows 10+. There is no Linux build.

Try it in Aiola: Download Aiola free for macOS 12+ or Windows 10+. Use Pro or the seven-day trial for hosted Analytics, then review Google-GeminiNotebook in the AI view.

Common questions

More from the blog

Build, ship and fix in one app

Errors, user feedback, GitHub and analytics live next to your coding agents in one desktop app, so every issue becomes a task an agent can take.