Website and API agent-readiness audit

When this helps

You may want an audit because:

• Your pages, structured data and API give different answers about the same product.
• An automated client cannot find or read information that matters.
• You want to know what AI assistants currently say before deciding what to change.

One example: the price is wrong

In the fictional sample report, one product appears with a real price on the page, a zero price in its structured data and an empty price in the API.

An assistant may quote different information depending on which source it reads. Correcting those facts is the first priority in the sample, even though its scanner does not score the conflict.

See the evidence and the fix

What I check

The audit covers up to two targets named in the written scope, for example a Shopify store and a B2B site, at the one fixed price. Each hostname named in the written scope is one target: a site and its API on different hostnames are two targets, and on the same hostname they are one. The written scope splits the 20 recorded questions between them, and the scan and the re-scan run on each target.

Three kinds of evidence, kept apart in the report so that you can see what each check found.

Can an automated client find and read your information?

I use an independent scanner and review the files, metadata and access rules that affect discovery and access.

Do your published facts say the same thing?

I compare information across your pages, structured data and API. The report shows where they disagree and what to change.

What do AI assistants say about your product?

I ask 20 recorded questions across three assistants named in the written scope. The questions contain only the public product name and the question text, never client material. The synthetic sample report shows a different run with 15 questions across four assistants, because it only illustrates the format. You receive the answers, the date and the conditions of the test. Each question is asked once per assistant, and an assistant can answer differently from one day to the next, so a changed answer after the fixes is reported as an observation.

On a large site, structured data, prices and availability are read on every page reachable from the sitemap and the API, as the sample report does for all 138 of its product pages. Head metadata and response headers are read on a sample of page templates, and the report lists every page it read. If a tool quota limits the checks, I raise the quota rather than reduce coverage. The written scope records the hostnames and the access I need before work starts.

Technical coverage

The scan uses isitagentready.com on its default profile. I record each check and its result, together with the measurement date.

The manual review covers /.well-known/ files, JSON-LD, head metadata, HTTP headers, robots.txt, sitemap.xml, ai.txt, llms.txt and the consistency of published facts. I also read your published web security scans.

The AI questions test both what assistants say about your product by name and whether they mention it when asked about its category. Questions, assistants and test conditions are recorded separately from the technical checks.

What you receive

A written report that answers four questions:

• What did I find?
• What evidence supports it?
• What should your team fix first?
• How can the correction be checked?

You also receive the recorded AI questions and answers, one round of written follow-up questions submitted within 14 calendar days of the report and answered within five business days, and one re-scan within 30 days of the report, or within 30 days of the delivered corrections when the correction add-on is bought. The written scope agreed before kickoff sets how many questions that round covers.

Each finding includes correction instructions and, where available, a link to the relevant guide. The order reflects the effect on users and agent behaviour, with effort and scanner effects recorded separately.

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

We agree the work. Send the URLs of up to two targets and your question. I confirm the scope, price and start date in writing. The audit is invoiced when the report is delivered, and the card link pays that invoice.

I run the checks. I complete the scan, manual review and AI questions, recording the evidence and dates.

You receive the report. The findings and correction plan arrive within two weeks of the agreed written kickoff. One round of written follow-up questions is included. Submit it within 14 calendar days of the report and I answer it within five business days. The written scope agreed before kickoff sets how many questions that round covers.

After the corrections, you choose the day for the included re-scan within 30 days of the report, or within 30 days of the delivered corrections when the correction add-on is bought. Technical fixes are checked with the relevant scanner or a direct test. AI answers are observed again using the same question set. The follow-up shows both readings and any changes in the method.

If you want help with the fixes

Your team can implement the report. Alternatively, a €499 add-on covers its complete fix list when bought together with the audit, with its report, or after the report and before implementation starts, and when the required access is arranged in advance.

For this add-on, I need an edge runtime in front of your origin, deployment access and any other access the listed fixes require, such as DNS. Access is agreed in writing and limited to the work.

If those prerequisites cannot be arranged, the add-on is not sold. Your team still receives the correction instructions. Work outside the list is scoped separately at €1,500 per day. Before that work starts, I agree the completion deadline as a stated number of business days from kickoff, the same arrangement the services page states.

Implementation scope and access requirements.

Monthly measurement is available through the separate advisory service, which does not include continuous monitoring.

Corrected 2026-09-28. Named the monthly cadence and the continuous-monitoring exclusion.