Skip to content

AI outage triage

When a monitor with Investigate outages (AI) enabled goes down, Uptime Cloud runs a short background investigation and attaches a likely root cause to the incident. This never delays the initial “down” alert — the investigation runs after it and delivers its findings as a follow-up a few seconds later.

In parallel, it re-probes the failing URL (status, headers, body), does a DNS lookup, checks whether the underlying host provider (Cloudflare, Vercel, Netlify, GitHub, Fastly, Heroku) is reporting an incident, and looks at certificate/domain expiry. Then it combines two things:

  • Deterministic facts decide the category and headline with high confidence when the evidence is unambiguous — e.g. the domain has expired, the TLS certificate is invalid/expired, DNS doesn’t resolve (NXDOMAIN), the host provider has a major incident, the CDN can’t reach the origin (edge 5xx), or the page shows a database-connection error.
  • AI (Cloudflare Workers AI, a Llama model) turns those facts into a plain-English explanation and a concrete next step for you.

Facts always win; the AI narrates on top and never overrides the evidence. If AI is momentarily unavailable, you still get the deterministic result.

The incident on the monitor detail page shows a “Likely cause” callout — a one-line summary, a short explanation, a suggested next step, and the supporting evidence (DNS result, detected host, re-check status, etc.). You also receive it as a follow-up push notification and webhook titled “TRIAGE REPORT” (see push notifications and webhooks).