Page Audit, Feed Audit, and Markdown for Agents

The Technical section bundles three tools that improve how machines read you. They share a theme: AI systems don't experience your site like a browser does, and the gap is measurable.

Page Audit

Run an audit against any URL. The result scores four dimensions:

• Access controls — can AI crawlers reach the page at all (robots rules, bot protection, status codes)?
• Content delivery — does the page deliver its content in a machine-readable way (server-rendered text vs. JS-only, structured data)?
• Content quality — is the content specific and answer-shaped enough to be citable?
• Overall — the combined score.

Audits are kept in history, and results can be exported as an HTML report for stakeholders. Audits draw from a monthly quota that scales with plan.

Use it when: a page you care about never gets cited. The audit tells you whether the problem is access, delivery, or content — three very different fixes.

Feed Audit

Checks your product feed against the OpenAI Product Feed Spec — the format behind ChatGPT's product surfaces (what you see in Product Radar). It flags compliance issues and highlights the LLM-optimized title and description fields most feeds leave empty. Feed audits have their own small monthly quota.

Use it when: competitors' products appear in ChatGPT product cards and yours don't, or your surfaced prices/titles look wrong.

Markdown for Agents

Serves AI bots an optimized markdown version of your pages instead of full HTML — same content, up to ~80% fewer tokens. Lighter responses mean crawlers read more of your site per visit and models ingest your content with less noise. The page shows 30-day serving stats once enabled.

Use it when: you have long, template-heavy pages (docs, product catalogs) and want crawl depth and ingestion quality to improve without touching the human-facing site.

Related

• AXO tests whether an agent can operate your site; these tools test whether machines can read it. Do reading before operating.
• LLM Site Visibility shows whether the fixes work: crawl activity and error rates are the ground truth.