Research with your own sources: from a pile of files to an evidence table

Organize documents around a question, compare disagreements, and build a source-linked answer you can check.

A useful research output shows where the evidence runs out.

Ask a question the collection can answer

“Tell me everything about remote work” invites a broad summary. “What do these three staff surveys say about meeting load?” gives the collection a defined job. Inventory the sources first: author, date, scope, and limitations. A small staff survey and a company-wide policy document describe different things. Include that distinction before trying to combine them.

Extract before synthesizing

Start with a table containing claim, source, location, and qualification. Keep direct observations separate from your interpretation. For example, a survey may report that respondents dislike afternoon meetings; it does not show that every employee shares that preference. Once the evidence table is sound, ask for patterns and disagreements. This two-stage approach gives you something concrete to audit.

Keep contradictions visible

One report might count scheduled meetings while another counts attended meetings. Different totals could reflect different definitions rather than an error. Ask what would explain the disagreement and which evidence could resolve it. Do not average incompatible numbers simply to produce a clean answer. Sometimes the best output is a short list of unresolved questions.

Inspect the source-linked answer

Google’s source notebook documentation describes working with supplied material and citations. Whatever tool you choose, open important citations and read around the passage. Tables, scans, footnotes, and missing pages can affect what was extracted. Before sharing a synthesis, check its strongest claim and a claim you initially found surprising. Include the collection date and known gaps so the next reader understands its limits.

A worked example

Original teaching example, not a recorded model run.

Vague starting point: Summarize all these PDFs.

A more useful brief

Using only these three reports, compare how each defines meeting time. Build a table with the definition, period covered, reported result, source location, and limitations. Keep disagreements visible. Then suggest what we can responsibly conclude.

Why this helps: This creates an evidence trail and delays the synthesis until the underlying definitions are visible.

Your turn

Choose two short public documents on the same subject. Extract one agreement, one difference, and one unanswered question.

You are looking for this: Every conclusion points to a specific passage and preserves its scope.

A quick check

Two reports measure different things. Should you average their numbers?

  1. Yes, the average is more balanced
  2. Only if both numbers are large
  3. First establish whether the definitions and populations are compatible

Answer: First establish whether the definitions and populations are compatible

A clean-looking average can hide an invalid comparison.

Sources and how to read this guide

Examples and exercises are original teaching material, not recorded model runs, measured time savings, or certifications. Official sources were consulted September 6, 2026. Product behavior can change.

Take the idea further

All learning paths | Practice challenges.

Related Power of AI pages

Keep reading with Start here, everyday uses, the tool guide, the writing workshop, and sources and standards.