Inspect the page as chunks
Rendered content is split the way a retrieval system would read it. You can see whether the opening states the topic, and whether later sections still belong to it.
See what an AI system is likely to retrieve.
The crawl evidence behind every retrieval test
Chunks, similarity, and the source URL stay attached to every match — so you can see what a model would retrieve, and from where.
- Rendered content chunks
- The passages cut from the page after JavaScript has rendered it.
- Cosine similarity
- How close each passage sits to the query, not a keyword overlap score.
- The opening chunk
- Whether the first section actually states the page’s purpose.
- Headings as retrieval units
- Sections that have to stand on their own when a model reads them apart.
- Buried answers
- The useful sentence sitting too far down the page to surface first.
- Source URL for every match
- Every retrieved passage still points at the page it came from.
How the retrieval audit comes together
Read the page in the chunks a model would use, run a query against the crawl, then see why a passage rose — or why a better one sat too far down.
Search the site like a model
A natural-language query ranks passages by semantic similarity, not keyword overlap. Open any result to the matching excerpt and the URL it came from.
Explain why something failed to surface
Weak retrieval usually has a structural cause: a buried answer, a heading that cannot stand alone, or an opening that does not match the rest of the page. The finding stays on that passage.
An AI-readiness audit you can inspect.
Chunk-level inspection
See the sections an AI system can retrieve on their own.
Topical focus
Catch openings and sections that drift from the page purpose.
Traceable retrieval tests
Follow every semantic match back to its source page.
Related workflows
Get started
Talk to your first crawl.
Example questions include: Show me internal linking opportunities, Audit this site, Where should I start?, Find pages competing for the same intent, What changed since the last crawl?.
Crawl a site, then ask what needs attention.
free to start