Automating your research: faster reading, not faster believing.

Decide the question first.

Vague research produces vague summaries. Write the question, what decision it supports, and what would change your mind. This gives both you and the assistant something to measure the sources against.

Gather sources you can name.

Build a list of the actual documents, pages, and datasets you will use. Ask the assistant to summarize only those and to quote the lines it relied on. Summaries that cannot point back to a passage should be treated as suspect.

Extract into a table.

For comparing many sources, ask for a consistent set of fields: claim, evidence, date, source, confidence. A table shows gaps quickly: the empty cells are the work still to be done.

Verify what matters.

Assistants can sound certain about things that are wrong. Check any number, date, name, or quote that will end up in front of someone else, using the original source. Label what you verified and what you did not.

Save the process, not just the answer.

Keep your question, source list, and extraction template. Next time, the research that took an afternoon will take less, and a colleague can follow what you did. Our Sources & verification page shows how we handle this on this site.