What is Agentic Search?
Also called Agent-driven search
Definition
Agentic search is an information-seeking process in which an AI system chooses searches or tools, evaluates intermediate results, and adjusts its next actions. It describes an adaptive workflow rather than a single fixed retrieval request.
The operation of Agentic Search is explained.
An agentic implementation can search, identify a missing fact, and issue another query before producing an answer. The exact tools, limits, and degree of autonomy vary; the phrase does not identify one universal search algorithm.
The relevance of Agentic Search is explained.
A buyer’s assistant might compare software integrations, then investigate authentication support for the shortlisted products.
A vendor can support that investigation with accessible technical documentation and explicit constraints rather than unsupported category claims.
The evaluation of Agentic Search is explained.
Evaluation uses realistic multistep questions and available retrieval traces to check whether responses preserve source limitations. Verified evidence, task completion, and search counts are separate measurements.
Frequently asked questions about Agentic Search
Is query fan-out always agentic?
Multiple related queries alone do not establish that a system adaptively chooses its next actions.
Does agentic search always take action?
An adaptive search process can stop at producing information without executing a purchase or other external action.
Can vendors guarantee selection?
Vendors cannot guarantee how an independent agent will retrieve or evaluate their content.