For engineers
What changed in the audit pipeline and why. File paths, code references, the contract changes, the regression tests that hold the line.
Read the engineering viewMX
Machine Experience
MX Audit.
What an MX audit delivers, and how to verify it on your own machine.
An MX audit measures what AI agents and machines see on your site today, against the same standards human visitors expect. The deliverable is a tagged PDF carrying its own evidence chain, so a reader, a regulator, or an AI agent can verify the work without trusting our word for it.
Each audit travels with a provenance sidecar pair on disk, a Level 2 PDF/UA conformance declaration in the file itself, and the full AI evidence chain embedded inside the PDF's metadata. The same scoring vocabulary runs across every dimension. Findings written for the human reviewer ship in a separate sidecar so the client-facing report stays client-facing. The three reads below explain the deliverable from three angles. Pick the one that matches your seat.
Pick your read
For clients
What the report looks like in your hand and why it reads cleaner. One scoring language, a contents page on its own page, reviewer working set aside.
Read the client viewFor auditors
How an MX audit proves itself. Evidence chain inside the PDF, sidecar pair on disk, framework citations naming EU AI Act, EAA Directive 2019/882, NIST AI RMF, and others.
Read the auditor viewVerify it yourself
Every MX audit deliverable is built so any reader can verify the evidence chain without leaving their own machine. Drop a PDF into the inspector to read the embedded provenance and conformance signals; read the explainer to see what each signal means and why it is there.
MX PDF inspector
Drop any PDF into the inspector to read the embedded MX metadata, the AI provenance sidecar payload, and the PDF/UA conformance declaration. Runs in your browser; nothing uploads.
Open the inspectorWhat MX Compatible means
The explainer walks through the five signals an MX Compatible PDF carries: tagged structure, MX-namespaced XMP fields, conformance declaration, provenance sidecar pair, and embedded AI evidence chain.
Read the explainerWant an MX audit on your own site or content estate? Contact us or email info@cognovamx.com. The audit is the entry point; the discipline keeps the gain.