Google NotebookLM

Google NotebookLM fetches a web page when a person adds that URL as a notebook source. The saved source gives the assistant material it can summarize, answer questions about, and cite inside that notebook. This is a user-initiated retrieval rather than a scheduled crawl of the open web.

Google's source documentation says a web URL import takes the text from the specified HTML page. It does not import images, embedded videos, or pages linked from that URL, and paywalled pages are not supported. A PDF reached through a URL is handled as a PDF source instead.

Once imported, the page becomes one of the sources available to the notebook model. The user can select sources, request a summary, or ask questions that the model answers from that material. A fetch in server logs therefore means someone tried to bring the page into a research notebook; it does not mean Google NotebookLM is continuously indexing the rest of the site.

The request identifies itself with the token Google-NotebookLM. Its robots.txt behavior is partial because the fetch follows a direct user action and may bypass crawler directives. A matching disallow rule expresses the publisher's preference, but it should not be treated as an access-control boundary.

Being available to this fetch can let a page appear as a cited source in a NotebookLM response for the user who imported it. That is different from inclusion in a general AI search index or a change to Google Search ranking. The available documentation makes no such broader promise.

Google's source-import documentation does not say that a Google-NotebookLM fetch contributes the page to model training. It describes the page as source material for the notebook. The token is useful for log classification and robots.txt policy, but this bot provides no Web Bot Auth key directory or verified IP ranges for stronger request authentication.