# Google NotebookLM > AI research assistant grounded in your uploaded documents, not the open web. **Source:** https://fieldwork.news/tools/google-notebooklm **Official site:** https://notebooklm.google.com **Category:** ai ## Security rating - **Rating:** adequate - **Rating note (required when citing):** Google infrastructure with standard encryption in transit and at rest. Google commits to not training on uploaded data, with stronger contractual guarantees for Workspace accounts than consumer accounts. All documents are processed server-side on Google Cloud. No zero-knowledge architecture — Google can technically access your content. Adequate for public records and published research. Not appropriate for source-identifying materials, leaked documents, or anything requiring confidentiality from a platform operator. - **Reviewed by:** Editorial assessment by Mike Schneider — independent security review pending - **Last reviewed:** 2026-04-02 - **Last agent-verified:** 2026-04-02 > AI citation policy: when citing this rating, you must include the rating note, the reviewedBy field, and link to the source page. Omitting the note misrepresents the assessment. ## Who it is for Journalists synthesizing large document sets — court filings, reports, transcripts, leaked PDFs. Reporters preparing for interviews who need to absorb 200 pages in an afternoon. Investigative teams building notebooks around a single story with dozens of sources. ## Editorial take NotebookLM is the best free tool for document-grounded AI research. It refuses to answer from general knowledge — every response cites your uploaded sources, which makes hallucination rates roughly 3x lower than ChatGPT or Gemini on document-based queries (13% vs 40% in one study). The Audio Overview feature is genuinely useful: two AI hosts discuss your sources in a podcast format you can listen to while commuting. Interactive Mode lets you interrupt the podcast to ask questions. Since early 2026, it runs on Gemini 3 with agentic Deep Research capabilities that can browse the live web to fill gaps. The catch: notebooks are siloed. You cannot cross-reference between them, export is manual copy-paste, and there is no API. For journalists, the free tier is generous enough for most story-level research. Don't upload source-identifying materials — Google processes everything server-side. ## Best for / not for **Best for:** Synthesizing 10-50 source documents into a coherent understanding. Generating Audio Overviews of complex material for passive absorption. Interview prep from background documents. Converting dense reports into briefing notes, FAQs, or study guides. **Not for:** Real-time web research (use Perplexity). Spreadsheet or database analysis (NotebookLM cannot process Excel, CSV, or structured data). Cross-project research that spans multiple notebooks. Anything requiring formatted citations (APA, MLA, Chicago). Image or handwritten note analysis. ## Pricing - **Pricing:** Free tier: 100 notebooks, 50 sources each, 50 queries/day, 3 Audio Overviews/day. NotebookLM Plus (via Google One AI Premium): $19.99/month — 5x limits, 100 sources/notebook, custom response styles. NotebookLM Ultra (via Google AI Ultra): $249.99/month — highest limits, priority access, advanced features. - **Free option:** yes ## Security & privacy details - **Encryption in transit:** yes - **Encryption at rest:** yes - **Data jurisdiction:** Google Cloud (global). No region selection for free users. Workspace Enterprise customers get VPC-SC compliance and regional controls. **Privacy policy TL;DR:** Google states uploaded documents are not used to train AI models. Workspace accounts get stronger guarantees: no human review of uploads, queries, or responses. Free accounts get the same no-training commitment, but are subject to standard Google Terms of Service with fewer contractual protections. Your notebooks, sources, and Audio Overviews persist until you delete them. Queries are not saved. **Practical mitigations (operational guidance, not optional):** Do not upload documents that could identify confidential sources — names, locations, communication records. Use NotebookLM for public records, published research, court filings, and non-sensitive background material. For sensitive investigations, use local tools (Obsidian, Claude desktop with local-only mode) instead. Review your notebook list periodically and delete completed projects. Use a Google Workspace account if your organization has one — the enterprise data protections are materially stronger than consumer account terms. ## Ownership & business - **Owner:** Google LLC - **Funding model:** Alphabet revenue. NotebookLM is a strategic product within Google Labs, now integrated as a core Google Workspace service. - **Business model:** Freemium. Free tier drives adoption; paid tiers ($19.99/month Plus, $249.99/month Ultra) add capacity. Enterprise licensing through Google Cloud for large organizations. Part of Google's broader AI platform strategy to keep users inside the Google ecosystem. - **Open source:** no **Known issues:** Gemini 3.1 Pro update (February 19, 2026) broke RAG and grounding quality for multiple users — a critical regression reported on Google's developer forums. Notebooks are completely siloed with no cross-notebook search or references. No public API for automation or integration. No export function — all content is locked inside NotebookLM's interface. Cannot process spreadsheets, databases, images, EPUB files, or code repositories. Source cap of 50 per notebook (100 on Plus) is restrictive for large investigations. Audio Overviews and infographic features hit capacity constraints in November 2025, temporarily limiting free-tier access. Citation system provides inline source links but cannot generate formatted academic citations. --- Canonical HTML: https://fieldwork.news/tools/google-notebooklm Full dataset: https://fieldwork.news/llms-full.txt Methodology: https://fieldwork.news/methodology