What agent-readiness means

Agent-readiness is not a search ranking and not a marketing score. It is the answer to one question: when an AI shopping agent answers a buying question in your category, does it name you?

That splits into two independently measurable halves.

The two halves

Are you in the answer? Ask a model family a real shopper question and read what it says. Either your brand appears or it does not. This is measured by asking four model families five tailored questions and having a judge classify every response. It produces visibility and recommendation rate.

Can a machine read your site correctly? Independently of what any model currently says, does your site expose the structured facts an agent needs — what you sell, what it costs, whether it is available, what others say about it. This produces technical readiness and drives the missing-signal penalties.

The two halves fail independently, which is why they are reported separately. A brand can be well-marked-up and still invisible, or widely recommended with an unreadable site.

Why this is not SEO

Search returns a list; you can be tenth and still get the click. An agent returns an answer, frequently naming two or three options. Position eleven does not exist. The measurable unit is presence or absence in a generated response, not rank in a list.

The second difference is observability. You can see your search traffic. You cannot see what a model told a shopper about you last Tuesday — there is no referrer, no log, no console. The only way to know is to ask the models yourself, repeatedly, which is what an audit run is.

AEO and GEO

Both terms are in use, and the reports use them for the two halves:

Term	Covers	Reported as
AEO — answer engine optimization	Whether you appear in generated answers	visibility, recommendation rate
GEO — generative engine optimization	Whether your site is machine-legible	technical readiness

What the score is for

The score exists to be acted on, not admired. Every point it deducts is attached to a named cause — a specific question you were absent from, or a specific signal your site lacks — and every cause becomes a ranked fix. A score with no attached causes would be a number without a next step.

It should be read with its error bars. See Measurement reliability.