The rename from NotebookLM to Gemini Notebook also changed the crawler's user-agent. Because it's a user-triggered fetcher, it can pull pages from sites that try to block automated bots — so robots.txt won't stop it.
2026-07-19 · by Moe Ameen
When Google renamed NotebookLM to Gemini Notebook on July 16, 2026, it also renamed the tool's web crawler. The user-agent token changed from `Google-NotebookLM` to `Google-GeminiNotebook`, and Google marked the old token as deprecated. Reporting around the change says the old `Google-NotebookLM` string keeps working through a short transition — described as lasting until around August 2026 — after which anything you've hardcoded against the old name (in a firewall rule, an .htaccess file, or a robots.txt block) stops matching. Google's own crawlers-and-fetchers documentation now lists Gemini Notebook under `Google-GeminiNotebook` as the current token.
The reason this touches site owners is how the fetcher is classified. Google groups its automated agents into common crawlers, which obey robots.txt, and user-triggered fetchers, which do not. Gemini Notebook sits in the second group. As Google's documentation puts it, because the fetch was requested by a user, these fetchers generally ignore robots.txt rules. So a `Disallow` line aimed at the bot does not stop it — the tool can retrieve pages from sites that otherwise try to keep automated crawlers out. Search Engine Journal, which flagged the rename's scraping implications, notes the tool's Discover Sources feature fetches web articles on a user's behalf — it reported the feature pulling up to ten sources per topic, summarizing them, and sending no referral traffic back to the original page.
The fix is not robots.txt — it's blocking at the edge. Site owners who want to keep the fetcher out can add a firewall or web-application-firewall rule, or an .htaccess rule, that matches the `Google-GeminiNotebook` user-agent and returns a 403, ideally verified against Google's published crawler IP ranges so you're not blocking spoofers or the wrong agent. There's a real tradeoff to weigh first: this behavior is documented and intentional, not a policy violation, and blocking it also means the tool won't fetch your page when one of your own readers references it. Treat the specific dates and the ten-source figure as a snapshot — confirm the current token and behavior on Google's crawler documentation.
The uncomfortable read on this news is about visibility, not just security: tools like Gemini Notebook increasingly consume your work, summarize it, and send zero referral traffic back. Whether you block the fetcher or not, the lesson is the same — reach you don't own is fragile, and the durable move is to publish your own point of view, in your own voice, natively on the platforms where your audience actually is, instead of hoping an AI tool cites you. That is exactly what Kompozy is built to do: take one source — a report, a transcript, a topic — and generate a Blog Article, an Email Newsletter, a captioned Persona or Clipped Short, a brand-exact Carousel, Quote Graphics, and native Text Posts, all in your brand voice through the Persona Brief, then schedule and publish across all nine connected platforms from one queue. You're not renting attention from a summarizer; you're owning distribution across surfaces a fetcher can't intermediate.
There's also a fast, high-intent content play in the story itself. "Does robots.txt block AI scrapers?" and "how do I block Google-GeminiNotebook?" are queries site owners and SEOs are searching this week, and most sites in your niche have no clear answer posted yet. Drop your take and the technical facts into Kompozy and it turns one explainer into a set: a blog post that ranks for the how-to query, a carousel that walks through the robots.txt-versus-firewall distinction, a short vertical clip, and platform-native posts for LinkedIn and X — published while the topic is still hot. Being early and correct on a change like this is how a single explainer becomes a week of content and a wave of the exact readers who care about it.
Yes. When Google renamed NotebookLM to Gemini Notebook on July 16, 2026, it renamed the crawler's user-agent from Google-NotebookLM to Google-GeminiNotebook. The old token is now deprecated but reportedly keeps working through a short transition period until around August 2026, after which only the new token matches. Google's crawler documentation lists Google-GeminiNotebook as the current name.
No. Google classifies Gemini Notebook as a user-triggered fetcher, and its own documentation states that because the fetch is requested by a user, these fetchers generally ignore robots.txt rules. A Disallow line will not block it — you need a server-level control instead.
Block it at the edge, not in robots.txt: add a firewall, WAF, or .htaccess rule that matches the Google-GeminiNotebook user-agent and returns a 403, ideally verified against Google's published crawler IP ranges so you are not blocking spoofers. Also update any existing rules that reference the old Google-NotebookLM token, since it is being retired. Note that blocking also stops the tool from fetching your page when your own readers reference it.
No — it is documented, intentional behavior. Google separates automated common crawlers (which obey robots.txt) from user-triggered fetchers like Gemini Notebook (which generally do not, because a person requested the fetch). It is not a policy violation, which is exactly why the control has to be a firewall or user-agent block rather than a crawl directive.