- Methodology
- Parameters
- Action-Oriented API
Action-Oriented API
stableCategory: ai-signals · Methodology v4.5
Does your site expose API endpoints that let an agent do something (search, submit, book, order) rather than only read content?
Signal Source
- Source
homepage HTML (anchors + inline method examples) + prefetched OpenAPI spec- Kind
- http_response
Score Bands
| Verdict | Condition |
|---|---|
| Pass | 3 or more distinct action-oriented endpoints are detected — via <a href> links or inline 'POST /path' examples in the HTML, or via OpenAPI paths whose path or summary/description names an action (search, submit, book, reserve, order, checkout, create, update, delete, apply, register, login, signup) |
| Partial | 1 or 2 distinct action endpoints are detected, or the signals are unclear/mixed |
| Fail | No action endpoints are detectable in the HTML or the spec |
Description
What this parameter measures
Does your site expose API endpoints that let an agent do something (search, submit, book, order) rather than only read content? That is what this parameter checks. friendly4AI collects candidate endpoints from three places: <a href> links in the homepage HTML whose path contains an action keyword (/search, /submit, /book, /reserve, /order, /checkout, /create, /update, /delete, /apply, /register, /login, /signup), inline method examples such as POST /api/submit in the page text, and, when a spec is prefetched, OpenAPI paths whose path or summary/description names one of those actions. Detected endpoints are de-duplicated before counting.
Why it matters for AI-readiness
Read-only access lets an assistant describe your site; action endpoints let it complete tasks for the user. When an agent can call /search, /book, or /order directly, it can run the whole workflow (find a result, reserve it, confirm it) instead of handing the user back to a manual form. Clear, well-named action endpoints are what turn your site from a source of information into a service an AI can operate on a customer's behalf.
How we score it
This is a gradient AI-Specific Signals parameter, and the scanner counts distinct action endpoints exactly as the rubric describes: three or more pass (100), one or two score partial (50), and none fails (0). Both HTML-derived endpoints and spec-derived endpoints feed the same de-duplicated set, so a documented OpenAPI path and a matching homepage link count once. This parameter is API-conditional: its weighted contribution counts only when an API is detected on the site (via api-discoverability); when no API is present its effective weight is zero and it is excluded from the score, so a content-only site is not penalized for lacking action endpoints.
How to fix common issues
- Expose at least three clearly named action endpoints — for example
/api/search,/api/book, and/api/submit. - Use action verbs in the path or the OpenAPI summary so the scanner recognizes the endpoint's purpose.
- Document each action in your OpenAPI spec with a descriptive
summaryanddescription. - If you only have one or two actions today, add the obvious next ones (a search endpoint and a contact/submit endpoint cover most sites).
- Re-scan after publishing to confirm three or more distinct actions are detected.
Version History
- Introduced
- v1
- Last changed
- v4.4
Key takeaways
- Signal: homepage HTML (anchors + inline method examples) + prefetched OpenAPI spec
- Category: AI-Specific Signals
- Passes when: 3 or more distinct action-oriented endpoints are detected — via <a href> link…