How recall records are matched
Last updated Oct 1, 2026. USDA FSIS records are not yet available in this search. View source coverage and dates.
Keep the agency facts separate from the matching rules.
Source records and dates
Every record preserves an agency identifier, official URL and retrieval date. FDA records without a recall number use the agency event identifier together with a product-derived identifier; they are marked as such in the data. Identical listing rows for one FDA announcement are combined. Separate agencies can describe the same underlying recall, so counts mean source records, not unique safety incidents.
Reported status
Ongoing, completed and terminated are copied from agency fields. An announcement without a termination field is shown as a published notice, with no inferred ongoing status. FDA warns that enforcement status can be stale. A missing result does not establish that a product is unaffected.
Products and medications
Product rules match phrases in product descriptions and structured product names, within allowed product types. Ingredient lists and hazard descriptions are excluded from product matching. Medication pages require a generic name or corroborated brand in the actual product description or title; harmonized names alone do not establish that an unrelated product belongs on a drug page. Medication groups are editorial browsing categories. Ingredient classifications are checked against NLM RxClass FDA-label evidence; acetaminophen also uses MedlinePlus. These browsing groups do not recommend a treatment. Matching does not establish the affected manufacturer or lot; verify those in the notice.
Distribution states
States are derived from the distribution text, rather than the firm's address. Nationwide US wording is mapped to the 50 states; Washington, DC is included when it is named. Ambiguous abbreviations are accepted only in a state-list position. Exclusions are respected. International-only wording is not turned into US nationwide distribution.
Freshness and failures
The site can check official feeds hourly and recheck the five-year backfill daily. It shows actual successful dates on the coverage page. A failed or incomplete feed does not replace the last successful source data. Only pages whose underlying records or archive memberships change are cleared. The web search uses stored data and makes no live upstream request.