Agent Traffic Is Active Demand, Not Passive Audience
Search traffic sends you a reader who may decide something later. Agent traffic sends you a system that is already executing a decision and is reading your surface to complete it. Treating the second as a variant of the first is the mistake this page is about.
The distinction
A search visitor is a person who might decide something. They arrive, read, leave, think, come back, ask a colleague, and possibly act weeks later somewhere you cannot see.
An agent visitor is a system in the middle of executing a decision. It was asked to compare options, shortlist vendors, check a claim, or complete a task, and it is reading your surface to finish that job. The decision happens at fetch time.
That is what makes agent traffic active demand rather than audience. Audience is attention that may convert. Active demand is a process already underway that your surface either participates in or does not.
Why the difference is not cosmetic
Three consequences follow, and each inverts a habit built for search.
Extraction outranks persuasion. A page written to move a human through a narrative may present an agent with layout, interstitials, and prose whose claims are implied rather than stated. What the agent cannot extract does not enter the comparison. The most persuasive page in the category contributes nothing to a decision made by a system that could not read it.
The window is one fetch. A human who finds your page confusing can persevere. An agent working through a list has a budget, and an expensive page is a page it partially reads or abandons. There is rarely a second visit in which you do better.
You will under-count it, structurally. Agent visits are hard to distinguish from scrapers and crawlers, frequently attributed to nothing, and increasingly absorbed by answer engines that read on the user's behalf and never send a visit at all. The surface where a decision was influenced can show no traffic. Measuring agent demand by conventional analytics tells you mostly about your instrumentation.
What this asks of a surface
It asks for readiness before rhetoric — and readiness has a specific meaning: whether your surface can be found, understood, trusted, and acted on by an agent, across discoverability, legibility, trust, callability.
The first two are mechanical and largely solved; making a surface agent-legible is a short and checkable list. The third is where most surfaces are silently weakest: an agent completing a decision has no way to check what you asserted, so it either takes the claim on faith or discounts it. Neither is what you wanted. That gap is the subject of agent readability is not agent trust.
None of this is a reason to write for machines instead of people. The same discipline that makes a surface extractable — explicit claims, stated scope, structure that matches meaning — makes it clearer to humans. The failure mode is not writing for both. It is writing for neither while believing the traffic numbers.
Where this sits
Agent traffic as active demand is the commercial face of agent readiness. Readiness is a property of your surface; active demand is why the property has consequences now rather than eventually. For the larger object both belong to, see the Verse.
FAQ
- How is agent traffic different from search traffic?
- Search sends a human who may act later, elsewhere, after more research. An agent arrives mid-execution: it is comparing, shortlisting, or completing a task on someone's behalf, and it will act on what it can extract in that fetch. The decision happens at fetch time rather than days later.
- Does that mean agent traffic is more valuable per visit?
- It means it is differently shaped. Agent visits are fewer, less attributable, and much closer to a decision. Conventional analytics under-counts them and cannot easily tell an agent from a scraper, so the surface where the decision happened often looks like it received no traffic at all.
- What does an operator do about it?
- Stop optimising only for persuasion and start ensuring extraction. What an agent cannot find, parse, or verify does not enter the comparison it is running. Readiness precedes rhetoric here — a persuasive page an agent cannot read contributes nothing to a decision made by that agent.