Every image in the archive was fingerprinted with a perceptual hash. These are the sets that came back near-identical — the same picture appearing more than once, often under a different filename.
What this is: a mechanical, offline pixel-match — no AI, no external service. What it isn't: proof of wrongdoing. Organisations legitimately reuse their own photos, and “near-identical” is a similarity threshold, not a guaranteed exact match. Treat each set as context to weigh, not a verdict. How this works & its limits →