Methodology & data sources

Exactly where the records on this site come from, how they are processed, and where the gaps are. We would rather show the seams than hide them.

Sources

openFDA drug label bulk download (export of October 29, 2026) — the FDA's published label repository, the same label content served by DailyMed (US National Library of Medicine). Each label's package display panel carries the product characteristics we use: imprint code, colour, shape, size and score.
FDA NDC Directory (product file) — used as a joining backbone for product and labeler names. The NDC Directory itself contains no appearance data.

Labels for non-solid products (liquids, creams, gels, inhalers, injections, patches), products with no imprint, and veterinary products are excluded.
Imprints are normalised (upper-cased, separators removed) for searching, while the label's own spelling is kept for display — including its mistakes, which we never silently "correct".
Records are deduplicated on imprint + medicine + strength + form family, because the same pill appears under dozens of repackager package codes. The current database holds 3,826 distinct pills across 3,594 imprint codes.
Quality check: a sample of imprints from the finished database was compared against an established identifier's published pages and agreed on medicine + strength in every case checked.

Coverage. In the openFDA export, product characteristics survive only inside labels' display-panel text, so many marketed products never appear. 3,826 pills is a floor, not the full US market; the database grows as extraction improves and labels refresh.
Labeler names. The labeler (the company that makes or distributes the pill) is known for about 32% of records (1,210 of 3,826). Where it is unknown we simply do not show a labeler — we never guess one.
No photographs. There is no free, reliable, current photo library of pills — the main public set (NLM RxImage) was retired in 2021 over mismatches — so we identify by recorded characteristics rather than risk showing the wrong picture.
Label data is labeler-submitted. The FDA does not verify every field of every label. Colours and shapes are recorded as the label states them, oddities included.
History. Discontinued, foreign and compounded products are largely absent; a pill that is not found here is not proven to be anything in particular.

Data refreshes are rebuilt from new FDA exports on a rolling basis. Questions about the data: contact us.