About the Runexus public scanner
Runexus performs visitor-requested, bounded checks of public website information. A scan is a first look at available evidence, not a test of whether AI platforms recommend a business.
Scanner identity
Our stable user-agent is:
Runexus-AI-Presence-Scanner/1.0 (+https://runexus.io/scanner-info/)
A user-agent is not proof of identity. We do not currently publish dedicated outbound IP addresses; do not use this identifier alone to grant trusted access or bypass security controls.
What we request
The submitted public page, then robots.txt, sitemap.xml, llms.txt, ai-agent.json and .well-known/ai-agent.json at its final origin. Up to four supported public document links advertised by a readable page may also be requested. If the page is inaccessible, only the five root documents at the submitted origin are checked.
We use GET requests, follow at most four redirects per resource, and check each target against public-network restrictions. Homepage reads are limited to 1 MB and nine seconds per request; document reads to 100 KB and 4.5 seconds. Redirects can extend the overall duration. We do not crawl sitemap entries or follow arbitrary site links.
Optional machine-answer checks
When a business-bound document declares the supported rnx-agent-v1 interface, we may send two fixed informational POST queries: “What is the business name?” and “Can you guarantee a ranking or recommendation result?” We check the response format, business association and explicit unsupported response. These checks do not establish accuracy, owner approval or AI recommendations.
These requests use no credentials or cookies and request no tools, actions or transactions. Only public HTTPS endpoints are eligible. We reject all redirects, limit each response to 32 KiB and allow at most ten additional seconds in total. Other interfaces remain untested.
Access and limitations
We read initial HTML without executing JavaScript. We do not sign in, solve CAPTCHAs, or bypass access controls. Robots.txt is collected as evidence; this visitor-requested diagnostic does not currently interpret its directives as a crawl permission decision. Site operators can block requests through their server or firewall controls.
Authority Quick Scan
When a business identity can be established and search access is configured, we send up to four bounded discovery requests to Google Search grounding through Vertex AI using the public business name, domain and relevant location. Up to 12 candidates are retained and six independent pages captured, with no deep crawling or review pagination. These public search terms are sent to the search provider; they are not stored in conversion events.
Authority captures allow at most two redirects, 250 KB and five seconds each, with three concurrent captures and a 25-second overall Authority deadline. Each redirect and DNS answer is checked; connections use the validated address. Unavailable or ambiguous evidence withholds the Authority number. A supported class requires a captured, exact-business source from the supported publisher registry; snippets do not score.
Blocked, oversized and incomplete pages receive no homepage score. Available public documents can still be reported. Missing optional documents do not lower a score. A block against this scanner does not prove that other AI systems cannot access the website.
Diagnostic data
Scanner acquisition diagnostics contain a random request reference, a bounded outcome category, HTTP status, a generic challenge indicator, duration, redirect count and truncation flag. They exclude submitted domains, page content and raw provider errors. These operational logs are separate from conversion events; ordinary hosting request logs remain subject to the hosting configuration.