Two archives, queried live.

• The Met Collection API (Open Access)Rights field: isPublicDomain. 61 matched · 24 examined · 4 dropped by the rights gate. Kept only where isPublicDomain === true and a file URL was published.12 shown
• Smithsonian Open Access API (api.si.edu)Rights field: content.descriptiveNonRepeating.metadata_usage.access. 123467 matched · 40 examined · 17 dropped by the rights gate. Kept only where metadata_usage.access === "CC0" and an image resource was published.12 shown

These figures describe the pass run at 2026-09-01 21:25 UTC, during this request. A completed pass is kept for 24 hours and replayed for an identical search, which is what keeps repeated queries inside Smithsonian’s 30-requests-per-hour allowance. A pass in which either archive failed to answer is never kept — an outage must not be replayed as though it were a finding about the catalogue.

Wikimedia Commons and the Library of Congress are deliberately not wired. Wikimedia was confirmed, on a live query, to return CC-BY material inside the same result set as genuine public domain works, and filtering it safely has not been designed yet. loc.gov was blocked by bot detection during the source evaluation, so its rights vocabulary has never been observed — and nothing here is built on a field nobody has read.