How to read an agent readiness scanner report
A score is a summary of checks. Here is how to read past it.
Am I Ready for Agents? editors · 10 September 2026 · Reviewed September 2026
A scanner report lists checks your public pages pass or fail, usually rolled up into a score. Read the failed checks, not the number: each scanner uses its own list and scale, scanners work on public pages, and a score does not show whether customers' agents complete tasks.
What is in a typical report?
A score or grade, the categories it is built from, and a list of checks with pass or fail and, often, a recommendation. Cloudflare's Is It Agent Ready groups checks into discoverability, content accessibility, bot access control, protocol discovery and commerce. ora.ai grades from A+ to F across discovery, access, usability and payments, and is-agentic.com splits checks into essential, recommended and bonus.
Which four questions should you ask about any report?
- Which checks failed, and which of them affect our main task?
- Which checks do not apply to us? A recommended check for developer APIs may not matter to a restaurant.
- What could the scanner not see? Scanners work on public URLs; ora.ai, for example, states it cannot scan login-gated content.
- Does it test what an agent does, or only what files exist? Some tools, including ora.ai's Scan and Journey, run agents on the site; many check for files and settings.
Why do two scanners give different scores?
Because they check different things and weigh them differently. Neither score is wrong. Compare the failed checks, not the totals, and do not compare a scanner's score with your self-check level: one measures pages, the other your own answers.
Can a scanner replace the self-check?
No, and the reverse is also true. A scanner sees your public pages but not your decisions: who owns the agent policy, what your fraud team does with a purchase an agent starts, whether anyone has looked at agent payments. The self-check asks about those but cannot see your pages. Use both, and expect them to disagree in places; the disagreements are worth a question to your web team.
What should you do with the report?
Take the failed checks that affect your main task to your web team, alongside the matching questions. Re-run the scan after changes, and keep your own analytics as the measure of whether anything changed for customers.
Keep each report with its date. Scanners change their lists of checks as standards move, so a score from March and a score from September may not measure the same things. ora.ai, for example, says Deep Scan v2 separates Verified signals, backed by agent behaviour, from Emerging signals that are tracked but not yet scored.