Is your site ready for AI search?
SearchD fetches your public HTML with an AI-search crawler.Every result shows its evidence. Checks that need a separate engine test are never guessed.
Your raw HTML is the evidence
Up to 300 URLs from your sitemap, fetched with an AI search crawler's user agent. Measured page fields stay separate from Jev's excerpt assessments.
Ingestion to extraction
Five layers: crawl access, chunking, internal linking, schema graph, extraction readiness — each check carries its own evidence and its own remediation.
No invented citation rate
This audit runs no ChatGPT, Claude or Gemini query. Those checks return as unverified, and the score never counts them as won.
What the four scores measure
| Score | What it measures | Points it is built from |
|---|---|---|
| Crawlability | How much of the sitemap this audit could actually fetch, and whether AI crawlers are turned away before they get there. | Share of discovered URLs fetched, multiplied by 0.25 when robots.txt disallows an AI bot and by 0.5 when a root probe is refused. |
| Chunkability | Whether a fetched page breaks into passages a retriever can index separately. | Body text density 45, one H1 30, any HTML table 25. |
| Knowledge graph | Whether pages carry structured entity data, and whether it points anywhere an engine can resolve. | Parseable JSON-LD 60, absolute sameAs URLs 40. |
| Direct answer | Whether facts sit in a form that lifts out of the page whole. | Header-bearing fact tables 60, server-rendered text density 40. |
The overall score is the mean of the four. Checks this audit cannot measure without querying an engine are reported as unverified and score nothing either way.
Awaiting a domain
Playback of the collected HTML evidence. The crawl has finished.
- Pages fetched
- —
- URLs discovered
- —
- Evaluator
- —
- Cost, estimated
- —
- Audited at
- —
Site-wide checks
decided once for the domainNo response loaded.
Page evidence
per fetched URLNo response loaded.
Page matrix
One cell per fetched page. Pick the measured field to colour it by.
Hover a cell to read its path.
AEO technical audit
No response loaded yet.
Technical extraction readiness from fetched HTML. Not an observed AI citation rate.
Crawlability
Awaiting fetched evidence.
Chunkability
Awaiting fetched evidence.
Knowledge graph
Awaiting fetched evidence.
Direct answer
Awaiting fetched evidence.
Six measured page signals
- 1 · Canonicals
- 2 · One H1
- 3 · Heading order
- 4 · Context links
- 5 · JSON-LD
- 6 · Fact tables
Each axis is the percentage of pages with that recorded signal. These six signals do not change the overall score.
What this audit measured, and what it did not
Awaiting a response.
| Layer | Score | Passed | Needs action | Open checks | Evidence |
|---|---|---|---|---|---|
| No response loaded yet. | |||||
Page evidence
Awaiting a response.
Page matrix
One cell per fetched page, coloured by one measured field.
Hover a cell to read its path.
| Path | Chars | H1 | H1–H3 | Tables · semantic | Canonical | JSON-LD | sameAs | Links · in text |
|---|---|---|---|---|---|---|---|---|
| No response loaded yet. | ||||||||
Site-wide checks
Awaiting a response.
| Status | Layer | Check | Evidence | Remediation |
|---|---|---|---|---|
| No response loaded yet. | ||||
Hand the failing checks to your coding assistant
Built from the checks that did not pass, with the evidence string each one was decided from. Nothing is added that the audit did not measure.
Run an audit to build the prompt.
Priority engineering remediations
Awaiting a response.
Want SearchD to fix these defects?
SearchD builds verified AEO knowledge graphs, neutral comparative tables and extraction-ready pages, so the engines quote your facts rather than guessing at them. The call is 20 minutes, with Snow Lee.