AI research agent
An agent that takes a research question, searches multiple sources over multiple rounds, synthesizes a sourced report, and follows up with clarifications. Distinct from a search box.
A research agent does what a junior analyst would do: read the question carefully, plan the angles, search for each angle, evaluate what was found, ask follow-up questions, and write up a cited report. Output is usually 1,000–5,000 words with citations.
In 2026 the category leaders are OpenAI Deep Research, Gemini Deep Research, Perplexity Labs, and Elicit for academic work. They differ in source coverage (open web vs. paywalled academic), report style, and tolerance for ambiguity.
Research agents are unusually good for executive briefs, competitive intelligence, due diligence, and literature reviews. The failure mode is shallow consensus — the agent finds the popular answer and stops. Mitigate by asking for dissenting views explicitly.
Where this shows up
Frequently asked
How do I get a research agent to find dissenting evidence?+
Ask explicitly. Add "find the strongest argument against this thesis" or "what would a critic of this view say." Default behavior is to synthesize consensus.
Can I trust the citations from a research agent?+
Verify the load-bearing ones. Modern research agents hallucinate citations far less than they used to, but ~5% of cited claims still misquote or invent the source. Spot-check before forwarding.
