How squirrel checks external links
Learn how local and cloud external-link checks work, what an unverified result means, and when a failed link counts as broken.
How the check works
The local checker requests each distinct external URL with a browser-like set of HTTP headers. It tries HEAD, then GET when needed, and follows safe HTTP and HTTPS redirects. It does not send your site’s cookies or authorization headers to another site. A link that answers 404 or 410, points to a domain confirmed not to exist, or fails its TLS connection is reported as broken.
The cloud link checker uses Perch to check links from its shared cache and retry requests that a plain client could not verify. Sign in with squirrel auth login to run an audit with cloud link checking.
What unverified means
An unverified result means the checker could not determine whether a person can open the destination. Sites may refuse automated requests with bot protection, rate limits, a sign-in wall or a temporary network error. Cloudflare and Akamai error responses can also come from an edge service rather than the destination site. These links are informational and do not affect the score.
The audit reports a link as possibly blocked when a response commonly used to refuse automated requests appears, including 401, 403, 406, 429, 999, or a Cloudflare or Akamai 52x response. A provider’s challenge fingerprint also makes the result unverified. A timeout is unverified too. Other transient errors can appear as possibly broken warnings.
Why a 404 can need two failures
The cloud checker does not call a link broken from one failed check. A first 404 waits for a second check; only a confirming failure makes it a broken link. This avoids scoring a brief network or edge failure as a dead destination. While it waits, the report lists the link as awaiting confirmation. This is the two-failures rule tracked in #2148.
See broken external links for the check names, messages and configuration.