Website and API agent-readiness audit

When this audit is useful

The audit fits when one of these questions is open. It records what is there today and assumes nothing about your site in advance.

• Your published product facts may disagree. A page, its structured data and an API can publish conflicting facts, and an assistant may encounter different answers depending on which source it uses.
• An automated client cannot find or read content or an API that matters to you, and you want the evidence of what it receives today.
• You need a documented baseline of what AI assistants answer about your product before deciding what to change.

A finding that changes the fix order

In the synthetic sample report, the invented company Northwind Fasteners Oy publishes one product three ways.

• Visible product page: A real price and lead time
• Structured product data: €0.00 · InStock
• Product API: Empty price · Marked purchasable

In this sample, the scanner does not score the conflict between the product page, structured data and API, and it is still the first fix in the report. Conflicting product facts can lead an assistant to quote the wrong price or availability, so correcting those facts takes priority over improving scanner coverage. The scored gaps, markdown, llms.txt, headers and discovery files, follow in the same plan with their own acceptance tests. Fictional company and readings.

Inspect the finding in the sample report

What the audit covers

Three kinds of evidence, kept apart in the report because they fail differently.

Technical scan

An independent scanner, isitagentready.com, runs against the site or API on its default profile, and every check it runs is recorded one by one rather than as one headline number. The level moves with the check set the scanner runs on the day, so the report names the checks that failed and what each one costs to fix, and leaves the headline number out of it.

Manual review

Review of /.well-known/ manifests, JSON-LD, head metadata and HTTP headers. Review of robots.txt, sitemap.xml, ai.txt and llms.txt against current agent norms. Whether a page, its structured data and its API publish the same facts, a conflict the scanner in the sample report does not score. Your published web security scans are read alongside the agent checks, so the report rests on measurements you can re-run yourself.

Observed AI answers

A documented question set put to several AI assistants, recording what they answer about the site or API today and whether they name it when asked about its category rather than by name. The questions, the assistants and the conditions are written down so the run can be repeated.

Large sites are covered in full. If a site is big enough that the live checks reach a tool quota, the quota is raised rather than the coverage reduced.

What you receive

• Recorded technical checks and manual findings, with the evidence behind each issue.
• A documented question set and the answers observed across selected AI assistants.
• A correction plan ordered by impact, with instructions and acceptance checks.
• One round of written follow-up questions and one included re-scan within 30 days of the report.

Findings are prioritised by their impact on users and agent behaviour, with implementation effort and scanner effects recorded separately. Every finding carries a fix instruction and, where this site has a guide for that surface, a link to it, so your team can do the work without buying implementation.

What you do not get

• Calls or meetings
• Implementation of the fixes, which is optional and described below
• Ongoing monitoring, which is the advisory engagement on the services page

A synthetic sample report shows the format: the per-check scanner readings, the manual review, the AI visibility run, nine findings with their owners and acceptance tests, and the decisions the company makes. It is not a report on a real client.

How the two weeks work

Agree the scope in writing. Send the URL and the question the audit should answer. I confirm the scope, price and start date in writing, and no production credentials are needed.

Run the checks and document the findings. The scanner run, the manual review and the AI answer run happen inside the two weeks, each dated and recorded.

Deliver the report and the correction plan. Findings, evidence, priorities, instructions and acceptance checks arrive in writing, with one round of written follow-up questions.

The included re-scan is separate from the two weeks: one re-scan within 30 days of the report, on the day you name, after the corrections have been made.

Implementation is optional

Your team can implement the report. A €499 add-on covers the diagnosis's complete fix list when bought with the diagnosis and when the required access is arranged in advance. If those prerequisites cannot be arranged, the add-on is not sold and your team still receives the correction instructions. Work outside that list is scoped separately at €1,500 per day.

For an audit, the add-on requires an edge runtime in front of your origin, deployment access and any other access the listed fixes need, such as DNS. The access is arranged in writing before the work starts and limited to the work. Implementation scope and access requirements.

Questions before you start