Our methodology

How we source, verify, and update every recall on this site — and what we will never do.

Recall Watch is an independent consumer-safety service. We are not a government agency and we are not affiliated with the FDA, USDA, CPSC, or CDC. We surface and personalize the public recall and outbreak notices those agencies publish — that's the entire product.

Where every recall on this site comes from

Every recall we publish is pulled directly from one of four U.S. government databases. We do not republish consumer rumours, lawsuit complaints, social-media reports, or unverified press releases.

  • FDA openFDA — food, drug, and device recalls. Includes Class I, II, and III classifications.
  • USDA FSIS — meat, poultry, and processed-egg recalls. Includes Class I, II, and III.
  • CPSC — consumer-product recalls (toys, appliances, baby gear, electronics, durable goods).
  • CDC — multistate foodborne and zoonotic outbreak investigations connected to recalled products.

Every recall detail page links directly to the original agency notice. If you want to verify a single line on our page, you're always one click away from the source of truth.

Update cadence

We hit the public source databases on the following schedule:

  • Recall ingest: hourly. New FDA / USDA / CPSC entries normally appear on this site within 60 minutes of the agency publishing them.
  • Outbreak signals (CDC): hourly during active multistate investigations.
  • Source re-sync: we re-fetch each recall up to 24 hours after ingest to capture amendments — agencies often add lot numbers, expanded distribution, or severity re-classifications a day or two after the initial release.

Free-tier users receive matched recall alerts within 48 hours of ingest. Pro and Family users receive matched alerts in real time (push + email) the moment a recall is ingested and matches their household profile.

How we use AI

Some content on this site is drafted with AI assistance. We are explicit about this because we think transparency matters more than appearances.

  • Recall summaries on the detail pages ("What's recalled", "Why it was recalled") are generated from the official agency release. Every factual statement traces back to a structured field in the source data — brand, lot numbers, affected states, reason for recall, recall date, severity.
  • Blog posts (Class I deep-dives, weekly roundups, brand history pages, evergreen guides) are drafted by Claude (Anthropic) from the same source data and held in a review window before publishing.
  • What we will not do: invent lot numbers, invent symptoms, invent affected states, attribute quotes to agency officials, or pretend a draft was written by a person.

Editorial review for high-stakes content

Class I recalls, drug recalls, and active outbreak guides are held in a 30-minute review window before publishing. Within that window we can revise, archive, or delay the post. Routine recall summaries and weekly roundups publish automatically once generated.

Drug recall pages additionally render a stronger "don't change your dose without consulting your clinician" callout — see our medication recalls for an example. We will never tell a reader to start, stop, or change a medication.

What we will never do

  • No medical advice. We summarize a hazard and link to the official notice. We never tell readers what medication to take, what symptom requires the ER, or what allergen exposure to treat at home. For any health question, we point to a clinician.
  • No selling your data. Your allergens, household profile, and email are used only for matching recalls to you and sending you the product. We never sell or rent the list. Read the privacy policy.
  • No spam. Every email is either a recall match you opted in for, a personal-account email (welcome, family invite), or a digest cadence you chose. One-click unsubscribe on every send.
  • No paid recall placement. Brands cannot pay to have their recalls de-listed, downplayed, or buried. The site reflects the public record exactly.

Corrections policy

If you see an inaccuracy in any recall summary or blog post — a wrong lot number, a misstated severity class, a missing state — email editorial@getrecalls.com. We aim to respond within two business days. Confirmed corrections are made within 24 hours and the page's "last updated" date is bumped.

We do not edit silently on substantive recall content. Major corrections are noted at the top of the affected page.

Contact & questions

For editorial questions: editorial@getrecalls.com
For product / billing questions: just reply to any email we send you.

See how the editorial team works →