Site Scan
An instant technical check of whether AI crawlers can read your site: the eight categories, how scoring works, the fix prompts, and the scan history.
Updated:
Before an engine can recommend you, it has to be able to read you. Site Scan checks that in a few seconds and gives you a graded report with a concrete fix for every failed item.
Running a scan
Pick one of your brand domains and hit Run scan. Customize lets you narrow the scan to a site type or to individual checks:
- All: every category.
- Content site: crawlability, discoverability, content structure, answerability, content accessibility, bot access control.
- API / agent site: crawlability, discoverability, bot access control, API and MCP discovery, commerce protocols.
Free accounts get 2 scans a day. Paid plans are unlimited.
What gets checked
| Category | Examples of checks |
|---|---|
| Crawlability and bot access | HTTPS, homepage returns 200, response under 1.5s, robots.txt, sitemap declared and reachable |
| AI discoverability | llms.txt, llms-full.txt, RSS or Atom feed, canonical URL, DNS _aid record, discovery Link headers |
| Content structure | Title and meta description length, Open Graph tags, single H1, heading hierarchy, semantic landmarks, viewport, favicon, lang, hreflang |
| Answerability and structured data | JSON-LD, FAQ content, text-to-markup ratio, a clear quotable first paragraph |
| Content accessibility | Markdown content negotiation, .md variants of pages |
| Bot access control | AI crawler rules in robots.txt, Content Signals directives, Web Bot Auth |
| API, auth and MCP | API catalog, OAuth discovery, auth.md, MCP server card, Agent Skills manifest, WebMCP, A2A agent card |
| Commerce protocols | x402, MPP, UCP, ACP |
The last three categories only matter if agents are meant to transact with your product. A content site that skips them is not doing anything wrong, which is exactly why the site-type presets exist.
Reading the report
Each category gets its own score, and each check is a pass, a warning or a fail with an explanation. Two things matter more than the number:
- Fails in crawlability are blocking. Nothing else helps if the crawler cannot reach the page.
- Fails in answerability are the ones that most often explain βwe rank on Google but AI never quotes usβ. A page that buries the answer at the bottom is hard to quote.
Fix prompts
Every check has a How to fix button, and the whole report can be copied at once. Each fix comes in two wordings:
- For LLM: paste it into your coding assistant and it will make the change.
- For developer: plain instructions for a person.
This is the intended way to use Site Scan if you do not write code yourself: run the scan, copy the full fix, hand it over.
History
Every scan is saved per brand. Expand a row to compare it with the previous score, so you can see whether a change actually moved anything. Switching brands shows only that brandβs scans.
Related: AI Visibility Playbook, Citations, Domains.
Did the content help you?