How to Use NotebookLM Data Tables

What NotebookLM Data Tables Does

NotebookLM is a free AI research tool from Google that lets you upload documents, ask questions, and generate structured outputs grounded in your own sources. The Data Tables feature, launched in December 2025, adds the ability to extract information from those sources into organized rows and columns using natural language prompts. Unlike the chat interface, which returns prose answers, Data Tables returns structured output you can export to Google Sheets with one click.

This matters for anyone who regularly pulls numbers, names, dates, or comparisons out of long documents by hand. Instead of reading a 40-page PDF and copying figures into a spreadsheet, you describe what you need and NotebookLM builds the table from your sources.

What You Need Before Starting

You need a Google account. NotebookLM is free for individual use, with Pro and Ultra tiers available through Google One AI Premium. Data Tables is available to all users as of early 2026. No software installation is required since NotebookLM runs entirely in the browser at notebooklm.google.com.

Supported source types include Google Docs, Google Slides, PDFs, web URLs, YouTube videos, audio files, and plain text. You can upload up to 50 sources per notebook, with each source supporting up to 25 million words.

Step 1: Create a Notebook and Upload Sources

Go to notebooklm.google.com and click "New Notebook." Give it a clear name tied to your project, such as "Q1 Sales Reports" or "Competitor Research." Then upload your sources by clicking the plus icon in the Sources panel on the left. You can drag and drop files or paste URLs.

For best results, keep your sources focused on a single topic or project. If you are comparing quarterly reports, upload all the relevant quarter PDFs into one notebook rather than mixing them with unrelated documents. NotebookLM builds its understanding from everything in the notebook, so unrelated sources can introduce noise into your tables.

Step 2: Open the Studio Panel and Select Data Tables

On the right side of the interface, you will see the Studio panel. This panel contains all of NotebookLM's output tools: Audio Overview, Study Guide, Briefing Doc, and Data Tables. Click "Data Tables" to open the table creation interface.

You will see a text field where you describe what table you want. NotebookLM also lets you select a preferred language for the output, which is useful if your sources are in one language but you need the table in another.

Step 3: Write a Clear Table Prompt

This is the step where most people get mediocre results. A vague prompt like "make a table from my data" will produce a generic summary. Instead, specify the exact columns you want, the type of information each column should contain, and any filtering or grouping criteria.

Here are prompt examples that produce usable tables:

For meeting notes: "Create a table with columns: Action Item, Owner, Priority (High/Medium/Low), and Due Date. Extract all action items from the uploaded meeting transcripts."

For competitor research: "Build a comparison table with columns: Company Name, Product, Pricing Tier, Key Features (top 3), and Target Audience. Pull data from all sources in this notebook."

For financial documents: "Generate a table with columns: Quarter, Revenue, Operating Expenses, Net Income, and YoY Growth %. Extract from the uploaded annual reports for 2024 and 2025."

For survey results: "Create a table with columns: Respondent ID, Satisfaction Score (1-5), Top Complaint, and Suggested Improvement. Extract from the survey response documents."

The more specific your column definitions, the better the output. If a column requires a particular format (such as percentages or dates), state that in the prompt.

Step 4: Review and Refine the Table

After NotebookLM generates your table, review it before exporting. Check for these common issues:

Missing rows. If a source contains relevant data that did not appear in the table, the prompt may have been too narrow. Try broadening the criteria or explicitly mentioning the source.

Merged or confused data. Complex documents with nested tables or heavily formatted layouts can confuse the extraction. If your source PDF has tables within tables or merged cells, consider copying the relevant sections into a clean Google Doc and re-uploading.

Hallucinated values. While NotebookLM grounds its responses in your sources, edge cases exist. Cross-check any critical numbers against the original documents, especially financial figures or statistics you plan to share externally.

If the table is not right, you cannot edit it in place inside NotebookLM. Instead, adjust your prompt and generate a new table. Each generation is a fresh extraction from your sources.

Step 5: Export to Google Sheets

Click the export button on your completed table to send it directly to Google Sheets. The data lands in a new spreadsheet with columns intact and ready for formulas, charts, or sharing. You can also copy the table in Markdown format using the copy icon if you prefer to paste it into another tool.

Once in Google Sheets, you have the full range of spreadsheet capabilities: pivot tables, conditional formatting, VLOOKUP, and chart generation. This is where the real analysis begins. NotebookLM handles the extraction; Sheets handles the computation and visualization.

Practical Workflow: Turning 12 Monthly Reports Into One Dashboard

Here is a concrete example of a full workflow. Suppose you have 12 monthly sales reports as PDFs, one per month. Each report contains regional breakdowns, product line performance, and customer acquisition numbers buried in prose paragraphs and embedded charts.

First, create a notebook called "2025 Monthly Sales" and upload all 12 PDFs. Then use this prompt:

"Create a table with columns: Month, Region, Total Revenue, Units Sold, New Customers, and Top Product. Extract monthly data from each report. One row per region per month."

NotebookLM will parse all 12 documents and assemble a single table. Export it to Google Sheets, then build a pivot table to see annual trends by region, or create a line chart showing revenue by month. What would have taken hours of manual extraction now takes about five minutes.

If you work with data regularly but prefer not to write code or configure tools, platforms like VSLZ AI offer a similar workflow where you upload a file and describe what you need in plain English to get charts and statistical analysis from a single prompt.

Tips for Better Results

Use high-quality source documents. Scanned PDFs with blurry text will produce poor extraction. If possible, use native digital documents or OCR-processed files.

Keep notebooks focused. A notebook with 50 unrelated sources will produce noisier tables than one with 10 tightly related documents on the same topic.

Specify data types in your prompt. If you want dates in YYYY-MM-DD format or currency values with two decimal places, say so explicitly.

Use the references column. NotebookLM includes a references column in generated tables that points back to the source document. Use this to verify any figure that looks unexpected.

Iterate on prompts. Your first prompt rarely produces the perfect table. Treat prompt writing as an iterative process. Adjust column names, add constraints, or split a complex table into two simpler ones.

Known Limitations

NotebookLM Data Tables cannot process data that is not in your uploaded sources. It does not connect to live databases, APIs, or external spreadsheets. The tool extracts and structures information that already exists in your documents.

Tables are not interactive inside NotebookLM. You cannot sort, filter, or edit cells. All post-processing happens after export to Google Sheets.

The references column does not link back to specific pages or paragraphs within your source documents. You get a source name reference but not a clickable citation to the exact location. For high-stakes analysis, manual verification remains necessary.

Complex PDF layouts with nested tables, rotated text, or multi-level headers can produce extraction errors. Simplifying the source format before upload improves accuracy.

Summary

NotebookLM Data Tables fills a specific gap: turning long, unstructured documents into clean spreadsheet data without manual copying. The workflow is straightforward. Upload sources, write a detailed prompt specifying your column structure, review the generated table, and export to Google Sheets. The quality of your output depends almost entirely on the quality of your prompt and the clarity of your source documents. For anyone who spends hours pulling data out of reports by hand, this feature removes the most tedious part of the process.