27 Sep 2026 · standards v1.4.1
Every figure below is counted from the listings we ruled, not sampled or estimated. A listing fails when it carries at least one Critical or High finding against the published standards. Categories are scanned in cycles, so each row carries its own date. Scanned 14-27 Sep 2026 on the 98 standards then in force (v1.3.1). The 11 added on 28 Sep 2026 were applied to every listing that day, so the rulings stand on v1.4.1, all 109 standards.
| Category | Listings ruled | Failing | Share failing | With a Critical | Cycle |
|---|---|---|---|---|---|
| Stress | 402 | 329 | 82% | 28 | 14-27 Sep 2026 |
| Gut and digestion | 321 | 260 | 81% | 54 | 14-27 Sep 2026 |
| Cognitive | 187 | 146 | 78% | 29 | 14-27 Sep 2026 |
| Sleep | 351 | 267 | 76% | 34 | 14-27 Sep 2026 |
| Joint | 368 | 267 | 73% | 37 | 14-27 Sep 2026 |
| Immune | 194 | 138 | 71% | 30 | 14-27 Sep 2026 |
| Energy | 183 | 125 | 68% | 28 | 14-27 Sep 2026 |
| Sports | 353 | 241 | 68% | 31 | 14-27 Sep 2026 |
2,243 listings across 793 brands in total. Every listing is a best seller in its category, which is the hardest test available: these are the pages with the most eyes and the most to lose.
Method
How a listing is ruled
Published copy is read phrase by phrase and each phrase is ruled against the standards in force at the time of the scan, currently 109 drawn from 19 published rulebooks. Each finding carries the rule it breaches and the source that rule comes from. Every standard is published individually so a ruling can be checked rather than taken on trust.
What "failing" means here
At least one Critical or High finding. Critical is the kind that can draw a warning letter or an automated takedown. High is a claim that cannot be substantiated as written. A listing with only Medium or Low findings is counted as fixable rather than failing.
Movement
Not published yet, and not estimated. Until 16 September 2026 a rescan overwrote the verdict before it, so no comparable pair exists for anything already published. Every scan is now recorded separately, so the first before-and-after arrives with the next cycle. When it does, it will say how many listings improved, how many got worse, and how many did nothing.
What it does not measure
Product quality, formulation, testing, or whether a brand is honest. It measures published marketing language against published rules. A brand can fail this index and sell an excellent product, and the reverse is just as possible. The full list of what this does not tell you is published.
Use it
Citation
Claims Verified category claims-risk index, 27 Sep 2026, standards v1.4.1, https://claimsverified.org/research/risk-index/
Cite freely with attribution and the index date. The index is machine-readable at /research/risk-index.json, and the standards behind it at /standards/registry.json.
The full research page · Check one listing against the same standards, free
The independent standard for supplement and telehealth claims. Our AI rules a brand’s public supplement claims against 109 published standards. Each failing line is rewritten to the rules. A licensed pharmacist signs every Claims Review Report.