Topic: Google NotebookLM export

Google NotebookLM Export — What You Can Save and What You Can't

Google NotebookLM is not Google Gemini. Where Gemini is a conversational AI assistant whose sessions land in Google Takeout, NotebookLM is a document synthesis tool: you upload sources — PDFs, Google Docs, web URLs, YouTube videos — and it generates structured outputs like Briefing Docs, Study Guides, FAQs, and Audio Overviews. The critical portability gap: as of 2026, Google Takeout does not include NotebookLM notebooks or their generated outputs. Everything NotebookLM generates is ephemeral unless you manually copy it. For teams using NotebookLM for architecture research — vendor evaluation, technology comparison, RFC synthesis — understanding the export story before closing a notebook is the difference between a preserved decision record and a lost reasoning trail.

TL;DR

NotebookLM is not in Google Takeout. Generated outputs — Briefing Doc, Study Guide, FAQ, Timeline, Table of Contents — must be copied to Google Docs or a local file manually at session time. Notes (pinned AI responses) must be copied individually. Audio Overview MP3 files have a download button in the UI. The Q&A conversation thread has no export path. Source documents: Drive files and Google Docs remain in Drive when a notebook is deleted; directly uploaded PDFs are deleted with the notebook. The decision-capture workflow: save outputs to Drive as you generate them → download Audio Overview MP3 → copy Notes to a Markdown file → use the Briefing Doc as ADR Context material.

What NotebookLM actually is (and why it differs from Gemini)

NotebookLM is a document synthesis tool, not a conversational AI assistant. The distinction matters for understanding its export story:

This architectural difference explains why NotebookLM is not in Takeout: Takeout exports conversational data and account data. NotebookLM's generated outputs are derivatives of documents you already own (Drive files, PDFs you uploaded) — Google's model is that you preserve the source documents, not the derivative synthesis.

In practice, this distinction is cold comfort when you close a notebook and realize the Briefing Doc you spent twenty minutes reviewing is not recoverable without re-generating it from the same sources.

What NotebookLM generates and whether it's exportable

Output type Generated where Export path In Takeout?
Briefing Doc Notebook Guide panel Copy to clipboard or Save to Drive (Google Docs) No
Study Guide Notebook Guide panel Copy to clipboard or Save to Drive No
FAQ Notebook Guide panel Copy to clipboard or Save to Drive No
Timeline Notebook Guide panel Copy to clipboard or Save to Drive No
Table of Contents Notebook Guide panel Copy to clipboard or Save to Drive No
Audio Overview (Podcast) Notebook Guide panel — two-host AI conversation Download button in UI → MP3 file No
Notes (pinned AI responses) Notes panel (sidebar) Copy individually — no batch export No
Q&A conversation thread Main chat interface No structured export — copy-paste only No

The most significant gap is the Q&A conversation thread. When you use NotebookLM's chat interface to explore your sources — asking questions, getting cited answers — that conversation is not exportable as a structured document. Unlike ChatGPT or Claude sessions (which appear in the respective data exports as machine-parseable JSON), NotebookLM Q&A sessions have no data portability path beyond copy-pasting individual exchanges. For critical reasoning exchanges (a question that surfaced a non-obvious risk, a comparison that drove a decision), manual capture at session time is the only option.

Source document persistence

NotebookLM's data ownership model for source documents depends on how each source was added:

Google Drive files and Google Docs

When you add a Google Drive file or Google Doc as a NotebookLM source, NotebookLM reads the document but does not own it. The file remains in your Drive; deleting the notebook has no effect on the Drive file. The notebook stores a reference to the file, not a copy. If the Drive file is later deleted or moved outside the notebook's access scope, the notebook loses access to that source — but the file itself is unaffected by the notebook deletion.

Directly uploaded PDFs and files

PDFs and other files uploaded directly to NotebookLM (not added via Drive) are stored within the notebook. When the notebook is deleted, these files are deleted. Before deleting a notebook, review the sources panel: any source with a file icon that is not linked to a Drive file should be downloaded or moved to Drive if it is not already stored elsewhere. This is the most common source of permanent data loss when users delete notebooks they assume are safe to remove.

Web URLs and YouTube videos

Web URL and YouTube video sources are references only — NotebookLM reads them when the source is added (or when you refresh the source) but does not store a persistent copy of the content. Deleting the notebook has no effect on the original web pages or videos. However, if a web page changes or goes offline after you created the notebook, NotebookLM's view of that content may be outdated — there is no version-pinning of web sources.

Pasted text

Text pasted directly into NotebookLM as a "Copied text" source is stored within the notebook and is deleted with it. If the pasted content is important — an RFC excerpt, a vendor pricing table, a competitor feature list — save a copy to Drive or a local file before deleting the notebook. Pasted text is also the source type most likely to contain unique information not available elsewhere.

Capture workflow for decision research sessions

The practical workflow for preserving NotebookLM work before closing a notebook:

  1. Generate and save outputs as you work, not at the end. Use "Save to Drive" for the Briefing Doc, Study Guide, and FAQ as you generate them. Drive saves are instant and the output becomes a permanent Google Doc — available even after the notebook is deleted. Don't batch all saves to the end of the session; if you close the tab accidentally, mid-session outputs are lost.
  2. Copy key Q&A exchanges to your Briefing Doc. If the conversation thread contains exchanges that surfaced non-obvious risks or narrowed down options, paste the relevant question-and-answer pairs into your Briefing Doc or a dedicated decision note while the session is active. The Q&A thread is the highest-signal content for decision-capture but has no batch export path.
  3. Download the Audio Overview MP3. If you generated an Audio Overview, use the download button in the Notebook Guide panel. The MP3 is replayable and useful for sharing research context with stakeholders or onboarding a new team member who wasn't in the research session.
  4. Export the Notes panel. If you used the Notes feature to pin AI responses during the session, copy them to a local Markdown file or Google Doc. There is no batch export for the Notes panel — each note must be copied individually.
  5. Check directly-uploaded sources before closing. Review the sources panel: any source with a file upload icon (not a Drive link, URL, or YouTube icon) is stored within the notebook. Download those files before deleting the notebook if they are not already stored elsewhere.

NotebookLM in the decision-capture chain

NotebookLM occupies a specific position in the research-to-decision workflow that maps to the ADR lifecycle:

  1. NotebookLM (research synthesis) — upload the relevant sources: vendor documentation, benchmark papers, competitor feature comparisons, RFC drafts, sections of your architecture documentation. Generate a Briefing Doc and FAQ to orient yourself. This output maps to the ADR Context section: it documents what was known about the problem space and the available options when the decision was made. Save the Briefing Doc to Drive at session time.
  2. ChatGPT or Claude (deliberation) — with the NotebookLM Briefing Doc as context, use a conversational AI session to work through trade-offs: which option fits the constraints, what are the non-obvious risks, what would have to be true for this decision to be wrong? This output maps to the ADR Decision and Consequences sections. This session is the one that is exportable — ChatGPT and Claude both produce machine-parseable conversation exports.
  3. WhyChose (extraction) — export the ChatGPT or Claude conversation and run the WhyChose extractor to produce a structured decision record: what was chosen, what was rejected, what trade-offs were accepted. This record becomes the structured core of the ADR, with the NotebookLM Briefing Doc linked as the research context from step 1.

The Briefing Doc is the most durable artifact NotebookLM produces for this workflow — saving it to Drive at session time ensures it is available as ADR Context material even after the notebook itself is closed or deleted.

Tool Primary use Export coverage Decision-capture role
NotebookLM Synthesize uploaded sources into structured outputs Manual only (copy / download); not in Takeout ADR Context — documents what was known about the problem space
Gemini Deep Research Multi-step web research with cited sources Full report text in Google Takeout Gemini Apps Activity ADR Context — web-sourced evidence for the problem statement
Perplexity Deep Research Multi-step web research with inline citations Partial via GDPR request; no native export ADR Context — evidence gathering before deliberation
ChatGPT / Claude Trade-off deliberation, constraint exploration Full conversation history via in-app data export ADR Decision + Consequences — the deliberation that produced the choice

Get early access to WhyChose

Related questions

Does Google Takeout include NotebookLM notebooks?

No. As of 2026, Google Takeout does not include NotebookLM notebooks or generated outputs. Takeout covers Gemini Apps Activity (personal Gemini conversations), Gmail, Drive, Calendar, Keep, Photos, and many other services — but not NotebookLM. All generated outputs must be manually captured at session time: text outputs via Copy or Save to Drive, Audio Overview via the MP3 download button. The most reliable approach is to use Save to Drive for text outputs as you generate them rather than waiting until the end of the session.

Can I export all generated outputs from a NotebookLM notebook at once?

There is no batch export. Each output type must be captured individually: Briefing Doc, Study Guide, FAQ, Timeline, and Table of Contents via Copy or Save to Drive; Audio Overview via the MP3 download button in the Notebook Guide panel; Notes by copying each individually from the Notes panel. The Q&A conversation thread has no structured export path — copy-paste only. Generate and save outputs throughout the session rather than waiting until you are done.

What happens to my uploaded sources if I delete a NotebookLM notebook?

Google Drive files and Google Docs remain in your Drive — NotebookLM reads them but does not own them. Directly uploaded PDFs are stored within the notebook and are deleted with it — download them from the sources panel before deleting if they are not stored elsewhere. Web URL and YouTube video sources are references only and are unaffected by notebook deletion. Pasted text sources are stored within the notebook and are deleted with it.

Is the Audio Overview transcript available in the export?

No. The Audio Overview generates a two-host MP3 conversation that is downloadable from the NotebookLM UI. The underlying script that the AI hosts read is not exposed as a text document in the interface or in any export path. To get the transcript, download the MP3 and run it through an audio transcription tool (Whisper, Google Speech-to-Text, or equivalent). Audio Overview transcripts make dense research summaries that are useful as ADR Context material — the extra transcription step is worth it for decisions that warranted an Audio Overview.

Further reading