The dataset, as a download

25 people, 465 sourced claims, spanning 348 documented products. Each row is one evidence entry exactly as it exists in the record: who it's about, which product, the claim as recorded, the class and confidence the record assigns it, the original source and its URL, and the two dates every entry carries, when it was retrieved and when it was last re-verified.

Version v2026-07-27 · dataset last re-verified 2026-07-27

Both files carry the same 465 rows, in the same order, with the license and citation written into the file itself.

• JSONapplication/json
• CSVtext/csv

Every claim is classified into one of six evidence classes, from strongest to weakest. The counts below are the same claims this download contains:

• She said it herself157 claims
• Her artist used it on her59 claims
• The brand documented it20 claims
• Reported by the press138 claims
• A likely match, unconfirmed10 claims
• Inside a paid relationship81 claims

Full definitions of each class are on the Method page. Before a row can exist at all, it has to pass the same mechanical checks that gate every build: a live source URL, a retrieved date and a lastVerified date, and a confidence rating of high, medium, or low, under rules the checker enforces rather than trusts, a visual match can never be rated high confidence, and a high-confidence rating requires the claim's origin to have been traced. Any quoted text is capped at 25 words, so a claim is described in the record's own words rather than stretched by a long lift.

41 widely repeated claims that could not be sourced at all are kept out of this dataset entirely and published instead on the Corrections page.

This dataset is licensed Creative Commons Attribution 4.0 International (CC-BY 4.0). You may copy, redistribute, and adapt it for any purpose, including commercially, provided you credit Actually Worn, link back to this page, and indicate if changes were made.

• 2026-09-05

  First publication of this page and of the machine-readable download (actually-worn.json, actually-worn.csv). No claims changed; this documents the packaging of the existing records into a citable dataset, not a change to any record.

This log starts on the date this page was first published. It is not backfilled, no earlier packaging changes were tracked before this page existed.

Questions about this dataset, or a suspected error in a row, can be sent to record@actuallyworn.com. See Contact for how a reported error is handled.