Page evaluation - ai-cmo.dev

Page evaluation is a workspace diagnostic stored by the web application. It is useful for checking a submitted set of URLs now. It does not write the pipeline D4 audit or recalculate the SEO score.

What it runs

The worker launches desktop Lighthouse at a fixed 1440 by 900 viewport with these categories:

Performance
Accessibility
Best practices
SEO

It records LCP, CLS, TBT, and FCP with the category results.

Technical checks

The evaluation also records nine direct checks:

Check	Rule
`robots.txt`	File can be checked
`sitemap.xml`	File can be checked
`llms.txt`	File can be checked
Meta robots	Page directive
Canonical	Canonical link
Title	10 to 70 characters
Description	50 to 160 characters
HTTPS	Submitted page uses HTTPS
Viewport	Viewport metadata is present

Only one page-evaluation job can run per workspace. The list returns the latest 20 results.

Difference from D4 audits

Property	Page evaluation	D4 technical audit
Trigger	On demand from app or API	SEO pipeline
Storage	`app.sqlite3`	`geo.sqlite3`
Device	Local desktop Lighthouse	Mobile in the current web worker
Checks	9	11 frozen checks
URL selection	User-submitted array	Up to 21 canonical URLs
Score impact	Diagnostic only	Technical SEO component
Recommendation links	None in its contract	Failed checks can link to recommendations

Do not present a page-evaluation result as the persisted technical SEO component. Run the D4 audit and score pipeline for that claim.

Chat access

Chat can read a known result with get_page_evaluation. It cannot use the tool to cross workspace boundaries. The public API currently lists recent evaluations rather than exposing a separate single-result route.

​What it runs

​Technical checks

​Difference from D4 audits

​Chat access

What it runs

Technical checks

Difference from D4 audits

Chat access