HypeCheck Research · July 2026
We checked 2,623 supplement claims against the evidence
Every supplement claim is a chain with three links: an ingredient does X at some dose, the product contains that ingredient at some dose, and therefore the product delivers X. We traced that chain for every marketing claim on 633 supplement product pages — 2,623 claims in total, each checked against ingredient research and clinical dose ranges.
27% of claims fully held up. 11% had no credible evidence at all. And yet most products — 56% — still earned a passing verdict. The supplements are mostly fine. The marketing is the problem.
What 2,623 claims look like, sorted by evidence
Each claim got one of four verdicts. Definitions are in the methodology below.
- • 6 in 10 products make at least one exaggerated ("stretch") claim.
- • 28% of products make at least one claim with no credible evidence behind it.
- • The average product page carries 4.1 checkable marketing claims.
The same chain, four ways it can go
Creatine monohydrate is one of the best-studied supplements there is. The claim, the dose, and the research all line up. Every claim on this label checked out.
True — if you're magnesium-deficient. If you already get enough, the effect is minimal. The evidence is real but conditional, and the label doesn't mention the condition.
AG1 does contain all four categories — that part is true. But it likely underdoses most of them compared to standalone products. Containing a category is not the same as replacing it.
No third-party clinical trial backs this number. Customer satisfaction surveys are not evidence — especially when the doses behind them sit inside a proprietary blend.
Where the chain usually breaks: the dose
Most exaggerated claims don't lie about the ingredient. They lie about the amount. The studies behind an ingredient use a specific dose — and the product often contains less.
- • Where the label disclosed enough to check — 878 ingredient doses — 53% were below the dose used in the clinical research.
- • 1 in 5 products hides at least one dose inside a “proprietary blend,” which makes the check impossible by design.
- • 27% of products have at least one ingredient that is likely underdosed.
The ingredients themselves are better than you'd think
We rated the evidence behind 3,082 ingredient entries. Most ingredients are not snake oil: 57% have moderate or strong evidence for something. But only 1 in 4 clears the “strong” bar — and marketing rarely respects the difference.
What the markup pays for
For 476 products we could compare the price against equivalent single-ingredient alternatives. 43% carry a high or extreme markup — you're usually paying for the blend's brand, not the ingredients. The pattern across the dataset: simple, boring labels price honestly; kitchen-sink blends charge for the story.
Most products still pass
Here's the part a takedown site wouldn't publish: after checking every claim, 56% of these products earned a MOSTLY LEGIT or LEGITIMATE verdict. Creatine monohydrate, quality fish oil, basic vitamin D3 — products like these made every claim they could back up and no claims they couldn't. The industry's problem is not fake products. It's real products described dishonestly.
- • LEGITIMATE: 70 (11%) • MOSTLY LEGIT: 283 (45%)
- • OVERHYPED: 248 (39%) • MISLEADING: 29 (5%) • HIGH RISK: 3 (<1%)
How to use this the next time you shop
- Ignore the claim. Find the ingredient's studied dose, then check the label for the amount. If the math works, the claim probably does too.
- Treat “proprietary blend” as a no. It exists to stop you from doing step 1.
- Favor single-ingredient products with boring labels. In our data they pass far more often than 30-plus-ingredient blends.
Methodology
Data. Snapshot taken July 4, 2026: all 633 published product reviews on HypeCheck. Claims come from each product's official website, including label text read by vision AI. These numbers roll up from 633 live review pages you can check yourself.
Claim verdicts. Supported: clinical evidence backs the ingredient for this purpose at adequate doses. Partial: some evidence, but results vary or require specific conditions. Stretch: loosely related benefit, but the claim is exaggerated. Unsupported: no credible evidence the ingredient delivers this benefit. Verdicts are AI-assisted, checked against an ingredient knowledge base built from PubMed research, and gated by a publish-quality review. Full pipeline: how HypeCheck works.
What we counted (and what we left out). Where our data couldn't make a call, we excluded the entry instead of counting it against the product. The dose stat uses only the 878 ingredient doses where the label disclosed enough to compare against clinical ranges. The evidence stat uses only the 3,082 ingredient entries with a rating. The markup stat uses only the 476 products with a price comparison.
Limitations. Claims were collected from official product pages, not ads or social media — the real exaggeration rate in the wild is likely higher, not lower. Analyses are evidence audits, not medical advice. Where labels hide doses, we say “can't verify,” not “false.”
About the editor
Danny Kim — PhD Computer Engineering, University of Maryland. Day job: large-scale code-analysis systems at Endor Labs. Engineer by training, supplement-skeptic by experience. Not a doctor. HypeCheck analyses are evidence audits, not medical advice. More about the editor →
Buying a supplement this week?
Check its claims against the evidence first — it takes about a minute.
Check a product Browse 600+ reviews