How the catalog gets built.

Every device starts from openFDA — the FDA's public data set. We pull 510(k) clearances, recall classifications, and device descriptions on a recurring cadence. That's the "Listed" page; the manufacturer doesn't do anything for it to exist.

Curated details — descriptions, specs, IFUs, brochures, photos, videos — come from the manufacturer (Verified / Channel) or our editorial team working from public sources.

We organize the GI catalog into ten clinical categories (Endoscopes, Hemostasis, Resection, ERCP Accessories, Stents, Dilation, Tissue Sampling, Motility & pH, Access Devices, Accessories & Disposables) across 45 subcategories — by how clinicians research, not how FDA product codes are laid out.

New devices auto-categorize from FDA product code + name; edge cases are re-categorized by hand. See something in the wrong bucket? Flag it.

Each subcategory has a published JSON schema for the fields that matter — working length, channel diameter, field of view for an endoscope; loop size, wire gauge, mechanism for a snare. The schema is what makes side-by-side comparison work; without it, every PDF names the same thing differently.

Spec values are validated against the schema on entry. Wrong-shape or out-of-range values are rejected — we'd rather show nothing than wrong numbers.

Full policy: editorial integrity. In short: only NPI-verified (or international-equivalent) clinicians publish, every review carries a COI disclosure, onescratch — not manufacturers — pays reviewers, and reviews are moderated before publish.

No purchasing recommendations. No brokered contracts, group-purchasing, or sales commissions. No advertising dollars for review content or ranking. "Verified" means claimed by the manufacturer — not vetted by us.

Recall data comes from openFDA on the same cadence and is used to ground our AI agent. Because recalls match to a device at the product-line level — not the exact SKU — we surface them on request through the agent, with that context and a link to the FDA's record, rather than as a standalone verdict on the device page.

If we publish something wrong, we correct it visibly. Significant review changes after the 24-hour edit window get an editor's note; spec corrections are timestamped on the product page.