Agentic Search
Agentic search describes a search process an AI agent runs on your behalf: it decomposes the question, issues its own queries, reads results, and returns a conclusion rather than a list. No single originator can be established for the phrase.
What it actually means
In an ordinary search the human does the work: type a query, scan ten results, click, refine, repeat. In agentic search the model does that loop. It breaks a question into sub-questions, runs several searches, reads what comes back, discards what does not help, and composes an answer.
The practical consequence is that the query you optimise for may never be typed by anyone. A person asks a broad question; the agent generates a dozen narrow ones you never see. Your page competes in that hidden set.
Why the term is recorded as generic
There is no dated paper or first-person coinage behind agentic search. It is compound descriptive language, assembled independently by multiple people once agents began doing this, and the honest record is that nobody owns it.
This site records terms that way when the evidence supports it. Assigning an inventor to a phrase that plainly emerged in several places at once would make every other attribution on the site less trustworthy.
What it changes about the work
It raises the value of covering a topic completely rather than optimising one page for one phrase, because you cannot enumerate the sub-queries an agent will invent.
It also raises the value of being unambiguous. An agent discarding candidates at speed will drop a source it cannot quickly resolve, and it will not come back to reconsider.