Google quietly retired the Google-NotebookLM user agent in its user-triggered fetchers documentation, replacing it with Google-GeminiNotebook to match the NotebookLM to Gemini Notebook rebrand. Search Engine Journal’s Roger Montti reported the documentation change on July 18. The old string keeps working through August 2026. After that, it stops matching every hardcoded rule built around it. Site owners who never touched their configuration will not notice the change until traffic they thought was blocked starts arriving again.
The rename itself changes nothing about how the tool behaves. Gemini Notebook still lets users upload documents, pull in YouTube and audio sources, and generate research summaries grounded in that material. What changed is the identifier the fetcher presents in its HTTP requests, documented now with separate mobile and desktop agent strings plus the retirement notice for the legacy token.
The detail most teams will misread is what robots.txt was ever doing here. Gemini Notebook’s crawlers are user-triggered fetchers, meaning a person initiated the request by pasting a URL or running Discover Sources. Per Google’s own documentation, user-triggered fetchers do not obey robots.txt at all. A disallow rule aimed at Google-NotebookLM was never a control mechanism; it was a line of text the fetcher was free to ignore. The only enforcement points that ever worked are a firewall rule or an .htaccess directive matching the user-agent string, and those are exactly what breaks silently once the old token expires.
Discover Sources is the feature that makes this worth tracking closely. It can pull up to ten sources per query, generate an AI summary, and return that summary without sending a referral back to any source it used. A site can supply the material for ten separate answers and register zero clicks for it. That asymmetry, content consumed with no traffic returned, is the actual stake behind an otherwise routine documentation update.
Google also used the update to delete Project Mariner from the fetcher docs entirely. The retired agent, which had been cited as an example of a Google-hosted browsing agent, was scrubbed rather than marked deprecated. NotebookLM’s own listing was replaced outright rather than kept as a redirect notice, a pattern worth remembering the next time Google renames a product: old documentation gets removed, not archived.
Before August, run one audit: grep your robots.txt, firewall configuration, and .htaccess files for Google-NotebookLM and swap in Google-GeminiNotebook wherever it appears, while confirming the block sits in the firewall or .htaccess layer rather than robots.txt alone. Sites that relied on a disallow rule for Gemini Notebook were already exposed to Discover Sources scraping before this rename. The deadline just forces a check most should have run months ago.
Search Engine Journal’s Roger Montti reported the documentation change on July 18, 2026.
