Gemini Notebook (formerly NotebookLM) — source corpus
- Free allowance
- Also 500,000 words/source cap and no page limit.
- Verdict
- caveatusable, with a named limit or an unprobed read path
- Cost
- free tier
- Identity
- emailno card, no phone
- Binding limit
- per-object-capper-object-cap 200 MB
- Evidence
- reported6 sources; a third party said so
- Freshness
- stale on 2026-11-05checked 2026-10-21 (inherited); stale on 2026-11-05
| Free tier | Also 500,000 words/source cap and no page limit. unit TB; metered on unstated |
|---|---|
| Price | none stated |
| Limits |
per-object-cap 200 MB or 500,000 words. Notebook cap is per-user; source cap is per-notebook; 'sharing a notebook does not change the source limit for any collaborator'. Copy-protected PDFs are rejected outright. Consumer API is undocumented and Google has changed *enforced* limits without updating the published tables. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Identity | email email |
| Egress | No egress metering or fee anywhere in the product. Reads happen through Google's own UI/API, not a CDN URL. |
| Per-object cap | per-object-cap 200 MB (200000000 bytes) |
| Retention | not stated |
| Interface | rest, cli No official public API for consumer accounts. Working path is the reverse-engineered internal batchexecute API: notebooklm-py (19,838 stars, pushed 2026-10-21) exposes Python, CLI, MCP server, and a local REST server; notebooklm-mcp (202) and notebooklm-rest-api (91) wrap it. |
| Tooling | pip install notebooklm-py; notebooklm source add, notebooklm ask, notebooklm generate, notebooklm download. Google Cloud has a *separate* documented Enterprise API (notebooks.sources.uploadFile) that requires a Workspace/Cloud entitlement. |
| Composes with | 100 notebooks x 50 x 200 MB = 1 TB per Google account. Google accounts are free and email-only, so N accounts linearise. Upskilling to Plus/Pro only raises the per-notebook source cap (100/300), so the notebook count stays at 200/500 — the free tier already gives the widest notebook multiplier per dollar. |
| Reliability | READ-BACK IS EXTRACTED TEXT, NOT ORIGINAL BYTES. There is no 'download source' button; the platform keeps a static extracted copy and the original file is not retrievable. Third-party extensions (Notebook Clipper, nlm tools, notebooklm-tui) export sources as Markdown/TXT/PDF of the *extracted text*; notebooklm-tui saves PDFs as per-page PNGs. It is a genuine 1 TB of free ingest, and a genuinely usable grounded-RAG corpus, but it is not a byte-exact archive. Rate limits apply and heavy use is throttled; the shared CI account-pool discussion in the wrapper's own docs is the tell. |
| Parsed numbers |
|
Evidence 6
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/gemininotebook/answer/16269187?hl=en | 'The current limit is 500,000 words per source or up to 200MB for local uploads. There's no page limit.' |
| 2026-10-21 | support.google.com/gemininotebook/answer/16213268?hl=en | 'Notebooks | 100/user'; 'Sources | 50/notebook' (Standard column); 'Sharing a notebook does not change the source limit' |
| 2026-10-21 | github.com/teng-lin/notebooklm-py | 19,838 stars, description 'Full programmatic access to Gemini Notebook's features—including capabilities the web UI doesn't expose'; pushed_at 2026-10-21T19:38:11Z |
| 2026-10-21 | notebookclipper.com/docs/notebook/export-sources | export is 'the text Gemini Notebook extracted from each source, as Markdown, plain text or PDF' — not the original file |
| 2026-10-21 | ainotechou.com/notelm-source/ | '元のPDFには戻らない！HTML形式への変換' (does not return to the original PDF) |
| 2026-10-21 | docs.cloud.google.com/gemini/enterprise/notebooklm-enterpris | separate Enterprise API exists but is entitlement-gated |
Unknown values 3
$.per_object_cap.notesthe field may be stated in a source that was not read$.retention.notesthe field may be stated in a source that was not read$.retention.statedthe field may be stated in a source that was not read