Agent Buyability Check

Three different questions get confused for each other. SEO asks whether a search engine will rank you. GEO asks whether an answer engine will cite you. Agent readiness asks whether a machine acting for a buyer can act on what it finds. A page can rank well, be cited constantly, and still be impossible to buy from.

The check scores fifteen items across five weighted dimensions. Machine-readable content carries thirty points, because nothing downstream survives a page that renders empty. Evaluable facts carry twenty-five, most of it on published pricing. Action surface carries twenty, identity and trust fifteen, and agent access policy ten. An agent can work around weak trust signals; it cannot read a page that never renders, so the weights follow the order of consequence rather than the order of effort.

One failure overrides the total rather than subtracting from it. If content does not render without JavaScript, the score is reported as blocked, because the rest of it is theoretical until an agent can read the page at all. The test is view-source, not DevTools - both a browser and DevTools execute JavaScript, so a site can look complete in each and still be an empty div to a client that does not.

The most common gap is pricing behind a contact form: in AAJ's Agent Readiness Study 2026, only 81 of 295 YC B2B startups had a price in the raw HTML of their pricing page (the study treats that rate as a ceiling). To a person that reads as an invitation to talk. To a machine it is a missing value, and a missing value in a comparison means exclusion from the comparison - not flagged as interesting, simply absent, with a competitor's number returned instead. Publishing a price as text, with the unit and currency explicit, is the highest-weighted fix after rendering.

Answers of not sure are scored as misses and counted separately, because an agent does not extend the benefit of the doubt either. Those are the items worth verifying against the live site rather than assuming, and they are listed on their own for that reason. The same fifteen checks run as an AAJ agent skill, so an AI assistant can score a site directly.

All free tools

Run this in your agent

The free AAJ agent skill Agent Readiness Audit does this job inside a coding agent, on your own data. Install it with npx skills add sarojkjha/aaj-marketing-skills --skill agent-readiness-audit.

Using Claude or ChatGPT instead? Add the AAJ engines as a custom connector, with no install: https://aaj-engines-mcp.sarojkjha10.workers.dev/mcp