Dossier · Source-grounded workspace
Source-grounded reading workspace · last verified 2026-06-07
NotebookLM is the rare AI workspace where “show me the source” is the product shape. The card should celebrate that while making clear that grounded-in-your-corpus is not the same as exhaustive or bibliography-ready.
Google source-grounded notebook workspace for uploaded reading packs, grounded chat, citations, Audio Overviews, Flashcards, Studio artifacts, Source Discovery, Deep Research, Workspace access, and Enterprise governance.
Sixteen scored dimensions, grounded chat and Studio artifacts, Google AI tier matrix, Enterprise deployment notes, competitive lens, and desk rows waiting for your corpus benches.
List prices and pay-as-you-go entry points we can cite without running our own bench. Each tile links to a source when possible.
Limits & product surface
Non-price vendor claims — multipliers, caps, and API scope. Detailed matrices live in subscription and SDK sections below.
The public positioning for this product — the loop we score against on VerdictPal.
Upload PDFs, Docs, web captures, or YouTube transcripts — or run Deep Research to import web sources into the same notebook. Name sources in your methods.
Ask narrow questions; click every citation chip back to the passage. If two sources disagree, say which one wins in your notes — do not let recency pick for you.
Generate Audio Overviews, flashcards, or slides for learning — verify claims against the PDF before any graded submission.
Standard caps bite on Audio and Deep Research during exam week. Pro via Google AI Pro is the default desk pick for thesis-scale notebooks.
Click a tile for the editorial note. Color follows score: coral, yellow, mint.
Shape
Avg 76 · 62–88
Best-in-class for uploaded reading packs; not a reference manager or systematic review screener.
By group
One input box, many retrieval postures. Filter by tier.
Q&A constrained to uploaded sources with inline citation chips
Two-host podcast recap — Deep Dive, Brief, Debate formats; Interactive Mode on English audio
Narrated slideshow from corpus — daily caps vary by Google AI tier
Agentic web sweep → citation-backed report importable into notebook
Quick web source scan for manual import — lighter than Deep Research
Active-recall artifacts from your uploads — mobile-friendly
Visual topic maps and comparison tables exportable to Google Sheets
Veo 3 immersive video — Google AI Ultra only (English, 18+)
Public rows are vendor or third-party claims we logged with a date. Desk rows are reserved for VerdictPal self-run results.
| Measure | Result | Source |
|---|---|---|
| Sources per notebook (Standard) | 50 sources | notebooklm.google/plans |
| Sources per notebook (Pro) | 300 sources | notebooklm.google/plans |
| Sources per notebook (Ultra 30TB) | 600 sources | notebooklm.google/plans |
| Daily chat queries (Standard)Resets every 24 hours per Help Center. | 50 queries / day | Vendor quota matrix |
| Audio Overviews (Standard) | 3 generations / day | Vendor quota matrix |
| Audio Overviews (Pro) | 20 generations / day | Vendor quota matrix |
| Deep Research (Standard) | 10 runs / month | Vendor quota matrix |
| Deep Research (Pro) | 20 runs / day | Vendor quota matrix |
| Google AI Plus | 7.99 USD / mo | one.google.com |
| Google AI Pro | 19.99 USD / mo | one.google.com |
| Google AI Ultra 20TB | 99.99 USD / mo | one.google.com |
| Cinematic Video (Ultra 20TB)English only · 18+ · Google AI Ultra required. | 2 generations / day | Product rollout May 2026 |
Consumer caps from notebooklm.google/plans and vendor quota matrix May 2026. Columns map Standard (free) · Google AI Pro · Google AI Ultra 30TB. Plus sits between Standard and Pro on most rows.
| Capability | Free | Pro | Max |
|---|---|---|---|
| Sources per notebook | 50 | 300 | 600 |
| Daily chat queries | 50 | 500 | 5,000 |
| Audio Overviews / day | 3 | 20 | 200 |
| Video Overviews / day | 3 | 20 | 200 |
| Deep Research | 10 / month | 20 / day | 200 / day |
| Cinematic Video | — | — | 20 / day (30TB) |
| Notebooks | 100 | 500 | 500 |
| Watermark-free Studio | — | — | Yes (Ultra) |
NotebookLM has no standalone checkout — limits unlock through Google AI Plus, Pro, or Ultra (or qualifying Workspace / Cloud contracts). This block is consumer subscription pricing only.
Standard
$0
Google account
Pro
$19.99/mo
Google AI Pro · Student $9.99/mo US
Google AI Plus ($7.99/mo) doubles many Standard caps (100 sources, 6 Audio/day) — see notebooklm.google/plans. Workspace Business Standard+ includes NotebookLM at ~$14/user/mo.
No consumer developer REST API (unlike Sonar or Scira). Organizations deploy NotebookLM through Google Workspace or Google Cloud Enterprise with IAM, VPC Service Controls, and audit logging — procurement surface, not courseware endpoints.
Workspace
From $14/user/mo
Business Standard and above
Enterprise (Cloud)
~$9/license/mo
Google Cloud · 15-license minimum
Desk has not validated Enterprise SLA. For courseware integrations, use Gemini API separately — it is a different product surface with its own data terms.
Atlas bundle string · checked 2026-06-07
What we learned while building this dossier — not vendor copy.
Standard vs Plus vs Pro vs Ultra caps (sources, chats, Audio, Deep Research) live on the plans page bundled with Google AI subscriptions — not a separate NotebookLM checkout.
The former single Ultra tier became 20TB ($99.99/mo) and 30TB ($200/mo) plans with different Cinematic Video and Deep Research ceilings — syllabus should name the SKU, not just “Ultra.”
Desk pass found hosts dropping footnote caveats present in PDFs. Interactive Mode answers from corpus but does not fix oversimplification in the recorded script.
You can attach a NotebookLM notebook inside Gemini for general-assistant questions grounded in the same corpus — different UI, same trust rules on citations.
Where NotebookLM wins for cited research — and where a rival still belongs in the stack.
Perplexity discovers and cites the live web; NotebookLM analyzes the corpus you already trust. Pair them — Perplexity for source maps, NotebookLM for close reading — do not expect either to replace the other.
NotebookLM
Perplexity
Vendors named on the product about page — useful for procurement and privacy reviews.
Editorial lenses only. Subscription and API pricing live in their own sections above.
Lens 01
NotebookLM's contract is simple: answers come from sources in the notebook, with citation chips you can click. That is what separates it from general chat — and what breaks when your corpus is incomplete.
How we exercised the product. Step through each run before you trust the scores.
Test run
Step 1 of 5
Create a notebook and upload three course PDFs (methods, results, discussion chapters).
TASK
Ask 25 questions over a mixed reading pack with known answers and conflicting sources.
Pending.
↳ exact citation accuracy, quote support, conflict handling, refusal quality, missed caveats.
TASK
Generate Audio Overviews for five source packs and compare scripts to source caveats.
Pending.
↳ caveat retention, unsupported claims, pedagogy, listenability, correction burden.
TASK
Run Source Discovery / Deep Research and classify imported domains.
Pending.
↳ primary-source rate, weak-domain rate, duplicate sources, citation usefulness, import cleanup time.
TASK
Compare personal, Workspace, and Enterprise paths for a sensitive internal corpus.
Pending.
↳ admin controls, auditability, data region, VPC-SC/IAM, human-review/model-improvement boundary.
This tool is in the VerdictPal Citation Fidelity protocol — frozen question bank, dual-reviewer grading, results still pending. Citation Fidelity v0.1 →
Benchmarks that evaluate or report on this tool. Scores and caveats live on the benchmark page.
Personal NotebookLM lacks the compliance/admin features of Enterprise. Workspace admins can control access. Google Cloud upgrade/Enterprise docs state uploaded files remain within the GCP project, data regionalization is honored, and files/chats/model outputs won’t be reviewed by humans or used to improve generative AI models.
Composed workflows on VerdictPal that reference this card, not vendor marketing.
The model(s) that drive this card. The tool is the wrapper; the model is the engine.
Compare side-by-side → open compare
How this tool fits into a composed research stack:
The Pack · Editorial newsletter
One short email when a card ships or changes status. No tracking, no third-party analytics. Unsubscribe in one click.