Multi-step search
Multi-step search is a loop in which an agent uses the result of one search to decide what to search next.
What it means
Multi-step search is the core of deep research. A single search returns links, but a research question usually needs several rounds. An agent reads the results, notes what is missing, rewrites its query, and searches again. Each round can follow a lead from an earlier page, such as a company name, a regulator, or a dataset.
The number of steps is a trade-off. More steps can find harder sources and fill gaps, but they also take more time and can drift off the question if the agent follows weak leads. Good agents keep a plan, track which sub-questions are answered, and stop when the evidence is enough. When you evaluate a tool, ask how it decides to stop and whether it shows the steps it took.
Continue this reading path
Cite this entry
https://bestdeepresearchai.com/glossary/multi-step-search/