SEAFOOD RESEARCH SERIES | PAPER NO. 4
Scanning the Seafood Label: A Confidence-Graded Brand-Search Model
Most seafood sold in the United States does not carry an establishment number. FDA packaged-food rules let the label identify a marketer or distributor rather than the processor, so a consumer scan cannot, by default, produce a plant-level attribution. A brand-search model can still do real work — as a confidence-graded bridge from label to responsible company to public enforcement record — provided it never promises more than it can prove.1,2,5
Prepared for publication by Farm Animal Transparency with the assistance of AI | April 21, 2026
| “Brand search is a bridge from label text to responsible company, not a replacement for a plant-level lookup. Most seafood labels do not name the processor, and FDA rules do not require them to.” |
At a glance
| • Under FDA packaged-food rules, seafood labels must disclose a responsible firm but do not have to identify the manufacturing site; establishment-number lookups, standard on USDA meat and poultry, do not translate directly to FDA seafood. |
| • Shellfish and Siluriformes are the exceptions: shellfish tags carry a state-issued processor certification number, and catfish, basa, and swai carry a USDA FSIS establishment number inside the inspection legend. |
| • FAT’s brand-search design is organized as a confidence ladder — direct match, responsible-company match, probable brand association — each with explicit attribution language. |
| • FDA’s Food Traceability Rule is not enforced until July 20, 2028, and traceability lot codes are not required on consumer packaging, so label parsing remains the near-term path. |
Executive Summary
FAT’s existing seafood label evaluation routes catfish, basa, and swai through USDA FSIS establishment numbers. For every other species of seafood, there is no establishment number on the label. A brand-search model has to translate the text that is printed on the label into a tiered link to public enforcement records: FDA Enforcement Reports (openFDA), warning letters, import alerts, and the Interstate Certified Shellfish Shippers List. The value of such a model depends on how well it distinguishes between a direct identifier, a named responsible company, and a probable association — and on how faithfully that distinction is communicated to the user.1,3,4,5,6
Why Brand Search Is Needed
Federal labeling rules give FDA seafood a different structure than USDA meat and poultry. For seafood under FDA jurisdiction, 21 CFR Part 101 requires the label to show the “name and place of business of the manufacturer, packer, or distributor,” and allows the label to say “Manufactured for” or “Distributed by” when the named firm is not the manufacturer. The address shown may be the firm’s principal place of business rather than the actual manufacturing site, unless using that address would be misleading. That means a brand scan can often identify a responsible company, but not necessarily the actual processor.1
By contrast, USDA FSIS requires the round inspection legend bearing an establishment number on inspected meat, poultry, and — since 2016 — Siluriformes products. Molluscan shellfish operate under a separate regime in which processors certified by state shellfish control authorities appear on the FDA Interstate Certified Shellfish Shippers List (ICSSL), and shellfish tags and labels carry the certification number. For FAT, this means that seafood splits into three label regimes: USDA-inspected (Siluriformes), state-certified (molluscan shellfish), and FDA packaged-food (everything else). A brand-search model is built for the third case.2,3
Fields to Extract from the Label
A practical model captures three clusters of fields. Hard keys — UPC, shellfish certification number, FSIS establishment number (Siluriformes), and lot or batch codes — provide the best evidence for a direct match. Identity fields — brand name, common or usual name, and scientific name where present — support species-normalized searches against the FDA Seafood List. Responsible-party fields — the role token (“manufactured for,” “packed by,” “distributed by,” “imported by”), the named firm, and the address — anchor the attribution logic.1,4
Two further label fields, required for most retail seafood under Agricultural Marketing Service country-of-origin rules, are material: country of origin and production method (“farm-raised” or “wild-caught”). Because they are legally required disclosures, they are reliable disambiguators when a brand name returns multiple potential enforcement hits.7
Match Rules, in Three Tiers
The model evaluates three tiers and stops at the highest tier that fires.
Tier 1 — Direct match. A direct match fires when the label contains an identifier that resolves to a specific plant or to a specific enforcement record. Four inputs qualify: a UPC that appears in an openFDA Enforcement Report for a recall whose date and lot window include the package; a shellfish certification number that resolves to an ICSSL-listed firm; an FSIS establishment number that resolves to a Siluriformes establishment; or an explicit “Manufactured by [company]” statement where the named firm appears in an FDA enforcement record. In each case, the label text is the same identifier as the one in the record.2,3,5,6
Tier 2 — Responsible-company match. A responsible-company match fires when no direct identifier resolves, but the label names a “Distributed by,” “Manufactured for,” “Packed for,” or “Imported by” entity and the normalized firm name matches a recalling firm, a warning-letter addressee, or an active import-alert firm in FDA’s records. Normalization removes corporate-suffix noise (“Inc.,” “LLC,” “Co.”) and applies a conservative fuzzy threshold to account for spelling variants. This tier surfaces accurate company-to-record links while reserving plant-level attribution for Tier 1.1,6
Tier 3 — Probable brand association. A probable match fires when no named responsible company matches, but the brand text appears in the product-description field of an enforcement record and at least two of species, product form, country of origin, or retailer also match the label facets. Species matching is normalized through the FDA Seafood List, which resolves acceptable market names — recognizing “basa,” “swai,” and “tra” as Pangasius species and routing them to the Siluriformes lane, for example. The two-facet threshold reduces false positives from generic brand names.4,5,7
A fourth outcome — no direct link — returns advisory text rather than a match, and points the user at the specific label features that provide stronger evidence: the FSIS inspection legend for Siluriformes, the state shellfish tag for molluscan shellfish, and the UPC scan for active recalls on any seafood category.
User-Facing Wording
Copy discipline is the hardest part of a brand-search model. The goal is to be definite about what is known and explicit about what is not. FAT’s three tiers map to the disclosure vocabulary already in use elsewhere in the FAT App: known for confirmed, partial for a named responsible company, and a lower band for an unverified brand association.
A direct match is presented as a confirmed link: “This product is associated with [record] via [UPC / shellfish certification number / FSIS establishment number / explicit manufacturer statement]. The identifier on the label is the same one in the record.” A responsible-company match names the firm identified on the label, cites the records in which the firm appears, and carries the attribution caveat: “Under FDA labeling rules, the actual processor may be a different facility, and the address shown may be the firm’s principal place of business rather than the plant.” A probable brand association is presented as a lead, not a confirmation: “The brand [name] appears in [N] enforcement records where the product also matches on [facets]. The responsible company and the processor are not confirmed from the label alone.”1
Caveats and Scope
Three caveats apply. First, attribution: the model should not say “this product was made by X facility” unless the match is anchored by a hard key — a shellfish certification number, a lot or UPC recall match, or an explicit manufacturer statement. Otherwise the safer formulation is “this brand or labeled responsible company is associated with these public records.” Second, data coverage: the openFDA food-enforcement dataset does not expose a dedicated brand-name field; its searchable columns include product_description and recalling_firm, so brand matching is inferred from product text rather than pulled from a clean brand column. Warning letters and import alerts are more firm-centric and map to Tier 2. Third, traceability: FDA has stated that it is not enforcing the Food Traceability Rule before July 20, 2028, and the traceability lot code does not have to appear on consumer packaging. Brand and distributor parsing, not traceability-code lookup, is therefore the practical path for the next several years.5,6,8
How This Fits Inside FAT’s Scoring
This model extends FAT’s 15-category seafood scoring framework, particularly Category 13 (Enforcement & Compliance), which routes Siluriformes labels to FSIS and all other seafood labels to FDA registration-based lookups. The confidence ladder here becomes the operational detail inside Category 13 for those non-Siluriformes labels — the mechanics by which a disclosure status moves from missing to partial to known based on the strength of the underlying link to a public record.
Conclusion
A seafood brand-search model can materially improve consumer access to public enforcement records, especially for recalls, retailer private-label products, and shellfish. It works best when it is presented as a confidence-graded bridge — from label text to responsible company to public record — and when the attribution language adapts to the strength of the underlying evidence. It is not a substitute for a USDA-style establishment-number lookup, and it should not be described as one.1,2,3
Publishing Package
Suggested URL slug: seafood-research-series-4-scanning-seafood-labels-brand-search-model
Meta description: A confidence-graded brand-search model for the FAT App, built for FDA-regulated seafood where the label identifies a distributor or marketer rather than the actual processor.
Endnotes
1. 21 C.F.R. § 101.5, “Food; name and place of business of manufacturer, packer, or distributor.”
2. 9 C.F.R. Chapter III, Subchapter F; Food Safety and Inspection Service, “Mandatory Inspection of Fish of the Order Siluriformes and Products Derived From Such Fish,” Federal Register, Dec. 2, 2015.
3. U.S. Food and Drug Administration, “Interstate Certified Shellfish Shippers List,” updated periodically; National Shellfish Sanitation Program, Model Ordinance.
4. U.S. Food and Drug Administration, “The Seafood List — FDA Guide to Acceptable Market Names for Seafood Sold in Interstate Commerce.”
5. openFDA, “Food Enforcement Reports” dataset; U.S. Food and Drug Administration, “Recalls, Outbreaks & Emergencies.”
6. U.S. Food and Drug Administration, “Warning Letters” and “Import Alerts,” searchable by firm name and product.
7. Agricultural Marketing Service, Country of Origin Labeling (COOL) requirements for covered retail fish and shellfish commodities.
8. U.S. Food and Drug Administration, notice extending enforcement of the Food Traceability Final Rule (21 C.F.R. Part 1, Subpart S) to July 20, 2028.
