Research delegated to an agent
Deep Research is a feature where an AI autonomously searches, reads, and compares many sources on the web, then produces a cited research report. Gemini, ChatGPT, and Perplexity all offer something equivalent, and the name has become near-universal.
How it differs from ordinary AI search
Ordinary AI search looks at a handful of results and answers quickly. Deep Research spends minutes to tens of minutes working across dozens of sources, running additional searches when something needs corroborating, reconciling contradictions, and producing a structured long-form report.
Underneath, it behaves as an agent, combining a reasoning model with tool calls.
What it suits
- Comparing products or services on price, features, and reputation
- Building a first draft of market or industry research
- Background work for decisions with many conditions, such as travel or moving
- Getting an overview when entering an unfamiliar field
Using it well
The more specific you are about purpose, the angles you care about, and the options you want compared, the better the report. Since it can take a while, starting it and doing something else is the normal pattern.
Citations do not guarantee the quality of what was cited. For anything consequential, follow the links to primary sources. Research legwork by AI, verification by a person, is the working division of labor.
Implementations differ in name and strengths, so trying a few and picking the one that fits how you research is worthwhile. Turning an hour of background reading into ten minutes is one of the most tangible benefits AI offers.
Copyright and quotation considerations are the same as for any AI-generated text. Rather than publishing a report as-is, work it through your own words and checks.