How to use NotebookLM data tables for source comparisons

For NotebookLM data tables, define the columns you need, ask your notebook’s source-grounded chat to extract matching facts, and verify the result against the original material. Google’s current help calls the product Gemini Notebook; the cited documentation does not establish a dedicated Data Tables menu, so this guide uses a source-comparison request rather than an invented button sequence. fenbs can keep the verified findings and next research step outside one notebook.

Google describes Gemini Notebook↗ (opens in a new tab) as a research assistant built around sources and source-grounded answers. The important job here is extraction: bringing the same kind of fact from several sources into a form you can compare. A neat layout alone does not make the values comparable.

What should each row and column mean?

Choose one row unit before you write the request. A row might represent a research paper, a public proposal or one version of a specification. Avoid mixing an entire organization in one row with a single product in another unless that difference is intentional and clearly labeled.

Then define the columns in plain language. For a hypothetical comparison of public research papers, use title, publication date, research question, method, stated limitation and source reference. Decide whether a date means publication, revision or the period covered by the research. Different date meanings should have separate columns.

Keep interpretation separate from extraction. “The paper used interviews” is a source fact if the methods section says so. “This approach is unsuitable for our decision” is your judgment and needs its own explanation. For a learning-focused workflow, see studying with NotebookLM.

How do you prepare the sources for a fair comparison?

Use a small collection you are allowed to analyze. Identify the documents and versions in your own working notes before requesting a table. Start with material whose key facts you can check yourself, so you can tell whether the request is extracting evidence or merely producing a plausible summary.

1. Choose the question the comparison should answer, such as how several papers describe their methods.
2. Identify the sources you want compared and record their titles and versions.
3. Exclude unrelated material from this comparison and keep a list of what you left out.
4. Write the row definition, column definitions and missing-value rule before asking for extraction.

Google’s source guide↗ (opens in a new tab) documents adding material through Add sources. Its refresh rules distinguish Drive imports from other source types. Record which version you checked; do not assume an earlier comparison has updated just because an original document changed.

Preparing an input spreadsheet is a different task. The guide to NotebookLM XLSX sources covers that neighboring question. Here, the goal is to extract comparable facts after the relevant material is available.

What should you ask the notebook to extract?

Write a request that names the source set and defines the output. Ask for explicit missing values and references, then test the response on a small set of rows. Google’s chat guide↗ (opens in a new tab) describes asking questions about notebook sources; this example is wording to try, not a separate product command.

Compare only the named research papers in this notebook.
Use one row per paper.
Columns: title, publication date, research question, method, stated limitation, supporting source reference.
Use only facts stated in the sources.
Write "not stated" for missing information.
Do not merge conflicting statements into one answer.
Keep your evaluation outside the factual columns.
Flag any row you cannot support from the source material.

If the result combines several documents in one row, narrow the request to one document and check the extraction first. If a column contains a broad summary when you need a precise fact, rewrite that column definition. Change the question before adding more sources to an unclear comparison.

How can you verify the table before relying on it?

Check the facts that could change your conclusion first. Open the original material, find the supporting passage and compare the actual wording with the cell. A reference to the right document is only a starting point; the passage still needs to support the particular value you are using.

• Units: distinguish percentages, counts, durations and currencies rather than mixing them in one column.
• Scope: check whether a claim applies to the whole study or only one group.
• Dates: keep publication dates separate from the period described by the source.
• Missing values: preserve not stated instead of turning it into zero or no.
• Conflicts: retain both statements and identify their sources until a person resolves the difference.

Suppose one paper reports interviews and another reports a survey. You can compare those methods, but you should not imply their results measure the same thing without further review. Add a column explaining the difference if it matters to the decision.

Keep the verified conclusion and any unresolved cells with your saved work in fenbs. That prevents a later assistant from treating an attractive draft table as completed research.

What usually goes wrong with a source comparison?

A common problem is asking for a conclusion before defining the evidence. “Which proposal is best?” invites judgment before you have agreed what the rows measure. First extract the facts under clear headings. Then ask a separate evaluation question with criteria a person has chosen.

Another problem is losing the difference between the original source and your working copy. Keep a date and version beside the comparison. Google’s notes guide↗ (opens in a new tab) says changes in an exported Google Doc do not sync back to the original note; treat separate copies as separate records.

Do not fix a usage-limit problem by repeatedly submitting the same large request. Check the vendor’s current guidance through the NotebookLM limits guide, then continue with the smallest useful comparison. For a visual summary after verification, see notebook infographics.

How to keep checked research with fenbs

Keep the work so you can carry on from exactly where you left off, on any device and in any other AI. Save the question, the facts you checked and the cells that still need attention. The next session should know where to begin.

1. Create a free account.
2. Save what you were comparing, the column meanings and the source names you checked.
3. Write the conclusion you verified. List any missing or conflicting values separately.
4. Choose a supported AI from Integrations and ask it to read your saved work before continuing.
5. Before stopping, save the next source or cell to check.

The cited documentation does not establish a custom MCP connection inside Gemini Notebook itself. fenbs does not import your notebooks or replace your source files. If all you need is a one-time comparison inside a single notebook, keeping your review there may be sufficient. Browse the AI at work hub for other practical research workflows.