# Measure agent-readiness with evidence

A scan is one source of evidence. Combine it with direct technical checks and observed AI answers, and record the method and date behind every conclusion.

Agent-readiness is a property you can measure, so it should be measured rather than claimed. Three kinds of evidence answer three different questions, and a credible conclusion names which one it rests on. A technical scan reads what a defined set of checks finds on the live site at a point in time. A manual review reads what the scan does not score, the parts that need a person to look, such as whether a page's markdown twin actually matches its HTML or whether a manifest is internally consistent. Observed AI answers read what a named assistant says today when asked a buyer's question, which depends on that assistant's own retrieval and can change between sessions.

The difference between a scan and a self-assessment shows up the moment something changes. A header gets dropped in a deploy, or a manifest starts returning the wrong content type. A checklist filled in by hand still reads as done, because nobody re-ticked the box. A scan reads the live site and the category drops, which is the only signal that matches what an agent experiences. Neither replaces the other: a scan tells you what a defined check found, and a manual review catches what the check set does not cover.

## What a technical scan checks

A scan runs a fixed set of checks against the live site, such as whether llms.txt resolves, whether a sitemap lists the pages it should, and whether a response carries the content type it declares. Each check either passes or fails on the day it ran, against the version of the site live that day.

## What a manual review adds

A person reads the surfaces a scanner does not score: whether an example in a guide still matches the current API, whether a claim on one page contradicts a date or a number on another, whether a markdown twin has drifted from the HTML it is supposed to mirror. This is slower and it is where the reviewer's judgment does the work a fixed check cannot.

## What an observed AI answer shows

Asking a named assistant a buyer's question and recording what it says is a third kind of evidence, distinct from both of the above. It shows what that assistant retrieved and how it answered on that date, not what the site's technical readiness is, and it can change on a later run without the site changing at all.

## Recording the method and the date

Each conclusion should say which of the three it rests on and when it was taken. This is the standard turva.dev applies to its own site. An audit reports the exact checks that pass or fail, each failure comes with a concrete fix, and the categories the report named are the ones a later scan is checked against. Measured by an independent scanner, turva.dev reaches 100/100 and Level 5, Agent-Native, on isitagentready.com. Measured 2026-09-14. A later scan can read a different check set, so it is reported as a new measurement and never as a re-confirmation of the old one.

For an audit that reports measured results, naming the method behind each one, contact info@turva.dev.

## Frequently asked

**Why should agent-readiness be measured rather than asserted?**

A checklist filled in by hand records intentions. A technical scan records what a defined set of checks finds on the live site, and the two often disagree, especially after a deploy drops a header or changes a content type.

**What is the difference between a technical scan and a manual review?**

A scan runs a fixed set of checks and reports pass or fail against the live site on the day it ran. A manual review covers what the scan does not score, such as whether an example still matches the current API or whether a markdown twin has drifted from its page.

**Why can a site pass a checklist and still fail a scan?**

Because a checklist filled in by hand records intentions and a scan records what a defined check actually finds. The two often disagree, especially after a deploy drops a header or changes a content type.

## Sources

- [isitagentready.com](https://isitagentready.com/)
- [internet.nl](https://internet.nl/)
- [Hardenize](https://www.hardenize.com/)

## Related

- [What a website and API agent-readiness audit covers](/guides/agent-readiness-audit)
- [How to choose an agent-readiness audit](/guides/choosing-an-agent-readiness-audit)
- [Common agent-readiness gaps in a measured sample](/guides/agent-readiness-gaps)
- [Define what an agent may do with your data](/guides/letting-agents-act-on-data)