live public-page audit · read-only

Will an AI agent understand your website — and finish the job?

Scan the structure, controls, and trust signals that browser agents rely on. Get a prioritized upgrade queue with evidence, copy-paste code changes, and downloadable fix packs.

controlled e-commerce study
89.3%vs49.3%

strict task success on agent-ready vs baseline pages across 300 runs.

Read the paper ↗

Agent readiness scan

Paste a public website.

read-only · robots-aware
No login. No install. Up to six public pages.

Method · v1.0

Three dimensions. Twenty-two deterministic checks. Zero vibes.

The score operationalizes the paper’s framework into evidence a developer can inspect. It complements accessibility, SEO, and human testing; it does not claim to replace them.

01 · 40%

Interpretability

Semantic structure, headings, JSON-LD, entity exposure, alt text, stable links, and crawl signals.

02 · 35%

Executability

Explicit labels, native controls, stable IDs, machine-readable state, and observable task paths.

03 · 25%

Decision reliability

Freshness, provenance, evidence, policies, availability, and constraints required for a correct choice.

What the score means.

It measures whether useful signals are present and coherent in public markup. The cited study was a controlled e-commerce proof of concept, so the score is a remediation heuristic—not a guarantee that every model will behave correctly and not a legal compliance certification.

Open-Source Audit Tool

Complete scans. Full developer fix packs. 100% free.

Run full website audits instantly and download complete code patches, developer briefs, and JSON reports with zero payment or account setup.

Full Audit
All 22 deterministic checks across 3 dimensions
Developer Pack
Exact code patches, JSON report, markdown brief, and downloadable ZIP
Self-Hosted
Ready to deploy on Vercel out of the box