Placement observations
Unknown includes timeouts, rate limits, robots exclusions, challenges, login walls and pages a plain fetch cannot read. An unknown result is never a removal finding.
Observation and confirmed loss are different
One complete absent observation describes one check. Confirmed placement loss requires two complete absent observations at least thirty minutes apart. Unknown attempts preserve the last known observation and carry uncertainty; they cannot supply the evidence needed for loss confirmation. The local ledger keeps intent and observations separately. Its default CLI exit behavior also preserves unknowns: a successful exit alone does not prove every expected link was found.Destination health is independent
Destination checks inspect the exact target URL. Results usehealthy, unavailable and unknown, with HTTP status, redirect and timing context where available. A 404 or 410 observation can report unavailable; confirmed unavailable state requires repeated evidence under the thirty-minute rule.
A source page can still link to a destination that is unavailable. A healthy destination does not prove that a source page links to it. Consult both placement tools and destination tools when investigating a broken citation.
Contact and discovery evidence
Published contact extraction reports what saved public HTML explicitly contains. It does not infer an address, check deliverability or establish that a publisher wants your proposal. Source dates and limits travel with the result. See the contacts reference. Imported and discovered rows carry source provenance and coverage limits. A supplier’s observation date is distinct from a verification date. A row missing from an inventory does not prove a link is absent, and a competitor’s placement does not establish your eligibility. Use the capability status to check which research surface is available, then connect your agent for the relevant pilot tools.Download retained evidence
Read link history to find the observation identifier, then retrieve its retained evidence. The existing HTTP download route isGET /v1/observations/{observationId}/evidence; destination observations use GET /v1/target-observations/{observationId}/evidence. These routes return an attachment. HTTP setup explains authentication.
The shared evidence commands require their matching release. If the download returns 410 EVIDENCE_EXPIRED, retain the observation metadata and record the missing historical bytes. A new check can observe the current page but cannot recreate the old one.