Skip to content

Google-GeminiNotebook, Read Aloud and other Google fetchers

What triggers Google-GeminiNotebook, Google-Read-Aloud, FeedFetcher-Google and Google-Site-Verification in your logs, and what breaks if a firewall blocks each one.

By Crawlito · 8 min read

In short

  • Google-GeminiNotebook fetches individual URLs that Gemini Notebook users add as sources; Google's changelog renamed it from Google-NotebookLM on 16 July 2026, and the old token was supported until August 2026.
  • Google-Read-Aloud fetches pages for text-to-speech in Google Go, Google Read it and Read Aloud on the Google app, and the nopagereadaloud meta tag is the documented opt-out, not robots.txt.
  • FeedFetcher-Google polls RSS and Atom feeds for Google News and WebSub, ignores robots.txt, and Google says to serve it a 404 or 410 if you want it to stop.
  • Google-Site-Verification fetches Search Console verification files and tags, and Search Console keeps checking them, so blocking it can cost you verified ownership after a grace period.
  • Google says its user-triggered fetchers generally ignore robots.txt, so in practice a firewall or WAF rule is what stops them: block on purpose, not by accident.
On this page

Google-GeminiNotebook is the user agent Google uses when someone adds your URL as a source in a Gemini Notebook project. It fetches that one page at the user's request, does not crawl your site, and, like Google's other user-triggered fetchers, generally ignores robots.txt. It replaced the Google-NotebookLM token in July 2026. The same family includes Google-Read-Aloud, FeedFetcher-Google, Google-Site-Verification and a handful of smaller fetchers, and each one breaks something different when a firewall blocks it.

This post is a lookup reference for those fetchers, based on Google's user-triggered fetchers page and the pages it links to, read on 15 November 2026. Google-Agent, the browsing agent in the same list, has its own post: Google-Agent and user-triggered fetchers covers its strings, robots.txt status and Web Bot Auth signing. Google's crawlers that do respect robots.txt, such as GoogleOther, are covered in GoogleOther and Google-CloudVertexBot.

Google's user-triggered fetchers at a glance

Google describes its list as not exhaustive. At the time of writing it contains these entries:

Token in your logsGoogle's product nameWhat triggers a fetchIP file Google names
Google-GeminiNotebookGemini NotebookA user adds your URL as a source for a notebook projectNot stated
Google-NotebookLMFormer name of the same fetcherSame as above; supported until August 2026Not stated
Google-Read-AloudGoogle Read AloudA user with text-to-speech enabled visits your pageNot stated
FeedFetcher-GoogleFeedfetcherA user or service asks to display your RSS or Atom feed (Google News, WebSub)user-triggered-fetchers-google.json
Google-Site-VerificationGoogle Site VerifierSomeone verifies or keeps ownership of your site in Search ConsoleNot stated
GoogleProducerGoogle Publisher CenterA feed a publisher supplied for Google News landing pagesNot stated
Google-PinpointGoogle PinpointA user adds your URL to a personal document collectionNot stated
GoogleMessagesGoogle MessagesA URL is shared in a chat and needs a link previewNot stated
Google-CWSChrome Web StoreURLs a developer put in Chrome extension or theme metadataNot stated
Google-AgentGoogle-AgentA Google-hosted agent browsing or acting for a useruser-triggered-agents.json

Google's crawling changelog dates several of these additions: Google-NotebookLM on 9 October 2025, Google-CWS on 3 November 2025, Google-Pinpoint on 12 November 2025, Google Messages on 21 January 2026, and the rename to Google-GeminiNotebook on 16 July 2026, with the note that NotebookLM is now Gemini Notebook. The full list moves often, so check the changelog before you hard-code anything.

Google-GeminiNotebook (formerly Google-NotebookLM)

Google's Gemini Notebook entry says the fetcher requests individual URLs that Gemini Notebook users have added as sources for their projects. One request means one person pasted one of your URLs into a notebook. It is not discovery crawling, and Google's page does not say the fetched content feeds Search or model training.

Google publishes a mobile and a desktop string:

Mozilla/5.0 (Linux; Android 10; K) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/138.0.0.0 Mobile Safari/537.36 (compatible; Google-GeminiNotebook; +https://developers.google.com/crawling/docs/crawlers-fetchers/google-gemininotebook)
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)

Match on the token, not the whole string: Chrome version numbers in published examples are not a promise about real traffic. The link in the string now redirects to the Gemini Notebook entry on Google's user-triggered fetchers page.

If you block it: the user's attempt to add your page as a source is likely to fail, or to pick up your block page instead of your content. Google does not document what the notebook shows the user in that case. If you have old firewall rules or log filters that name Google-NotebookLM, update them: Google lists that token as supported until August 2026, so current traffic should carry Google-GeminiNotebook.

Google-Read-Aloud

Google's Read Aloud page says the fetcher lets Google Go, Google Read it, Read Aloud on the Google app and other Google text-to-speech services read web pages aloud. It is activated when a user with text-to-speech enabled visits a page. It does not follow links, and it renders pages statelessly, without the user's cookies. Google says it caches page results, so a user listening to a page may or may not cause a visit, depending on whether the page was fetched recently, though you may still see several requests for the same page.

The strings follow the same pattern as Gemini Notebook, with the token Google-Read-Aloud and a link to https://support.google.com/webmasters/answer/1061943. The older token google-speakr is deprecated.

Google says you cannot opt out with robots.txt. The documented control is a meta tag:

html
<meta name="google" content="nopagereadaloud">

Google notes that Read Aloud may proactively access and render the page, with its embedded resources, to recognize the tag; once it has seen the tag, future requests are reduced automatically. For paywalled articles, Google's advice is to mark the content with subscription and paywall structured data, with isAccessibleForFree set to false.

If you block it: visitors using those text-to-speech features cannot hear your page. Separately, because the fetch has no cookies, a page that only shows content to logged-in users is fetched as the logged-out version.

FeedFetcher-Google and GoogleProducer

Google's Feedfetcher page says Feedfetcher is how Google crawls RSS and Atom feeds, for Google News and WebSub. It fetches only the feed URL it was given and does not discover links. Google says it should not fetch feeds from most sites more than once an hour on average, with more frequent refreshes for sites that update often. Its string is short:

FeedFetcher-Google; (+http://www.google.com/feedfetcher.html)

Feedfetcher ignores robots.txt, and Google says it "behaves as a direct agent of the human user". To stop it, Google says to serve a 404, 410 or other error status to the Feedfetcher-Google user agent; for a feed on a blogging or hosting service, it says to work with that provider. Its IPs are in user-triggered-fetchers-google.json.

GoogleProducer is the Google Publisher Center fetcher. Google says it fetches feeds that publishers explicitly supplied for Google News landing pages.

If you block either: feed updates stop reaching Google News and WebSub, or the feeds you configured in Publisher Center cannot be read. One possible cause is a bot rule that treats short, non-browser user agents as suspicious. For how feeds help discovery in general, see RSS feeds and content discovery.

Google-Site-Verification

Google Site Verifier fetches the token you place for Search Console ownership. Its string is:

Mozilla/5.0 (compatible; Google-Site-Verification/1.0)

Search Console's verification help page sets the rules that matter here. The HTML verification file cannot require authentication, the homepage cannot require authentication for the meta tag method, and redirects to another domain are not followed for the file (the page allows same-domain redirects, such as http to https). It also says Search Console periodically checks that your token is still present and valid. If it cannot confirm it, you are notified, and if the issue is not fixed, your permissions expire after a grace period. The page says Google uses the Google Site Verifier user agent to perform verification.

If you block it: the first verification fails, and, more quietly, an existing file or tag verification can lapse later if the periodic check keeps being blocked. DNS verification uses a record at your domain provider rather than a file or tag on your web server, and it is the only way to verify a Domain property, so a Domain property is not exposed to this kind of block in the same way.

The smaller fetchers: Pinpoint, Messages and Chrome Web Store

Three entries are short in Google's documentation:

  • Google-Pinpoint requests URLs that users add to their personal document collections in Google Pinpoint.
  • GoogleMessages generates link previews for URLs sent in Google Messages chats. Blocking it is likely to leave a shared link without a preview. Other apps' preview bots are compared in link preview bots.
  • Google-CWS requests URLs that developers supply in Chrome extension and theme metadata. If you publish an extension, blocking it can stop the Chrome Web Store reading the pages you linked.

robots.txt and verification, in brief

Google says these fetchers generally ignore robots.txt rules because a user requested the fetch. A Disallow line under their token or under User-agent: * is not a dependable control. Where Google gives a specific control, it is something else: the nopagereadaloud tag for Read Aloud, an error status for Feedfetcher. For the others, a firewall rule or authentication is what actually stops them. The Google-Agent post covers this in more depth, and do AI crawlers respect robots.txt compares Google with other operators.

User agent strings can be spoofed, so check the source IP before you trust or block one. Google's verification guide lists the files for this family:

https://developers.google.com/static/crawling/ipranges/user-triggered-fetchers.json
https://developers.google.com/static/crawling/ipranges/user-triggered-fetchers-google.json
https://developers.google.com/static/crawling/ipranges/user-triggered-agents.json

Google says IPs in user-triggered-fetchers.json resolve to gae.googleusercontent.com hostnames and fetchers controlled by Google use user-triggered-fetchers-google.json and resolve to a google.com hostname. The user-triggered fetchers page words the same split by owner and pairs the patterns the other way round, so treat both ***-***-***-***.gae.googleusercontent.com and google-proxy-***-***-***-***.google.com as possible for this family. Since Google names the file for only Feedfetcher and Google-Agent, check an unknown fetcher against all three. A reverse lookup, then a forward lookup on the hostname it returns, gives a second check (the address below is a documentation example):

bash
# 1. Reverse lookup on the IP from your logs
host 203.0.113.10
# 2. Forward lookup on the hostname that step 1 returned;
#    it should resolve back to the same IP
host google-proxy-203-0-113-10.google.com

The verify real AI crawlers guide shows the same method for other operators.

How to find and fix accidental blocks

  1. Search edge logs, not just origin logs. A request your CDN or WAF rejects never reaches your server. Search security events for each token in the table above.
  2. Check the status codes. A 403, 429 or challenge page for an IP that is in Google's files points to a bot rule, rate limit or "unknown bot" setting.
  3. Allow by token and IP together. An allow rule that trusts the string alone lets spoofers through. Pair it with the published ranges and refresh them automatically.
  4. Decide per fetcher. You may want Google-Site-Verification and FeedFetcher-Google allowed while blocking others. Each is a separate token, so each can get its own rule.
  5. Retest the feature. Re-run Search Console verification, add one of your URLs to a test notebook, or check a feed in a reader, and confirm the request now returns 200.

Frequently asked questions

Is Google-GeminiNotebook the same as Google-NotebookLM?

Yes, it is the same fetcher under a new name. Google's changelog recorded the rename on 16 July 2026, and its user-triggered fetchers page lists Google-NotebookLM as the former token, supported until August 2026.

Can I block Google-GeminiNotebook with robots.txt?

Google says user-triggered fetchers generally ignore robots.txt, and its Gemini Notebook entry gives no robots.txt or blocking instructions of its own. A firewall rule or authentication is the dependable control.

How do I stop Google Read Aloud reading my pages?

Add <meta name="google" content="nopagereadaloud"> to the page. Google says robots.txt cannot be used to opt out of Read Aloud.

Why did my Search Console verification stop working?

Search Console periodically rechecks the verification token, notifies you if it cannot confirm it, and lets permissions expire after a grace period if the issue is not fixed. A firewall that started blocking Google-Site-Verification, a login wall or a redirect on the verification file are common causes.

Do these fetchers affect my Google Search rankings?

Google does not describe any of them as feeding Search indexing, which is Googlebot's job. Blocking them breaks the feature each one serves, so judge each by that feature.

Next step

If a security rule is catching Google's fetchers, it may be catching AI crawlers too. Run the free check on an important page to see whether the main AI crawlers can fetch it, with no account. It runs from Crawlito's servers, not Google's, so your edge logs remain the evidence for these specific fetchers.

Filed under AI crawlers. Topics: google, ai-crawlers, server-logs, robots-txt