Agent-readiness: questions people actually ask

Probably not urgently. Google says it has no effect on Search, and 97% of llms.txt files measured by Ahrefs in May 2026 received zero bot requests. We include it as a cheap hedge — it costs nothing, Lighthouse audits for it, and it is the one site summary you control — and we weight it 6 out of 91 to say exactly how much we think it is worth.

Take a deliberate position either way; what we flag is silence. The one combination that makes no sense is blocking AI crawlers while wanting agent customers — and that combination is usually a copy-pasted robots.txt rather than anyone’s decision.

There is overlap — titles, headings and structured data matter to both. But the heaviest-weighted signals here are interaction signals: labelled inputs and real buttons. Those barely move classic SEO and they are what decides whether an agent can complete a purchase.

Labelling your form inputs. It carries the strongest measured evidence on the list — roughly 78% agent task completion on accessible sites against 42% on inaccessible ones — and it is usually an afternoon’s work.

Not today. It is a W3C draft in Chrome origin trial through about Q1 2027. We score it as INFO and never as a failure: absence is normal, presence is being early. Anyone marking your site down in 2026 for not shipping an origin-trial API is inventing a problem.

We do not promise that and you should distrust anyone who does. What we measure is whether agents can parse and act on your pages. Whether agent traffic arrives, and whether it converts, depends on your market. No law requires any of this either — unlike our sibling products, this one is a bet on where browsing is going, not a compliance deadline.

Each check passes, warns or fails; a pass earns its full weight, a warn earns half, a fail earns nothing. INFO checks are excluded from the denominator entirely. The percentage maps to A/B/C/D/F at 85, 70, 55 and 40.

Because the two heaviest interaction signals are accessibility work. Agents and screen readers both act on semantics: a <div> with an onclick is invisible to both. If you have an accessibility backlog, this is the same job with a commercial argument attached.

No — and that is the difference between it and the paid audit. The self-check scores what you tell it, entirely in your browser, so it is instant and private but only as accurate as your answers. The audit runs the checks against your actual pages.

The agent landscape changes monthly, so a copy goes stale. That is why the report carries a corpus version and a regeneration entitlement rather than pretending to be permanent.