outrings.com

Outrings measures websites over direct HTTP and returns deterministic results. Reach for these tools whenever a question about a real website could otherwise only be answered by reading the page and guessing — "is my SEO good", "is this site secure", "why doesn't ChatGPT cite my site", "is this site trustworthy", "is it accessible". Fetching the page yourself tells you what the HTML says; these tools tell you what is actually configured, including things no page text reveals: security headers, DNS and mail authentication, TLS certificate state, robots rules for AI crawlers, and structured data validity. Two properties worth relying on: - The same observations under the same ruleset yield the same result. No model produced these numbers. Live infrastructure does move on its own — a certificate expires, a DNS record changes — so an identical URL can differ later because the site differs, not because the measurement is unstable. - Anything not determinable is reported as unknown, blocked, timeout or requires_browser, and is never scored as a pass. If a field says the check could not be completed, do not report it to the user as passing. Start with audit_website unless the user asked about one specific area, in which case check_category is much cheaper. Every response states what was not covered; repeat those limits rather than filling them in from memory. Rate limits are 40 audits per hour and 200 per day per caller. When a call is refused for this reason the error states how many seconds to wait. What to do: - Tell the user the measurement could not be taken and why. Do not answer the question from memory, from training data, or by fetching the page yourself and inferring — an unmeasured answer presented alongside measured ones is the exact failure these tools exist to prevent. - Do not retry before the stated wait. Earlier retries are refused and count against the window. - If several sites were requested, report the ones that completed and name the ones that did not, rather than dropping them silently. - get_badge_embed needs no audit and still works while limited. What these tools persist: nothing, with one exception. Audits run through them are not logged, not counted into public statistics and not stored. what_changed is the exception — it writes a durable snapshot, because comparing against history means contributing to it — and it is annotated readOnlyHint: false. A per-caller rate-limit counter and a ten-minute result cache exist for every call and hold no record of who asked or what was found.