About NutriVouch
A free reference that shows how much clinical research stands behind each supplement — and how little stands behind some of them.
Who makes this
NutriVouch is built and maintained by Denys Demchenko, a software developer. Not a doctor, not a nutritionist, not a researcher. The site does not produce medical opinions: it counts and links the published research, and the interpretation of each study stays with the people who wrote it.
That matters for how you should read this site. Where numbers come from a database query, they are exact. Where a page summarises what the evidence means, it is an editorial judgement — and every such page says so.
Why it exists
Most supplement pages are written by someone with a product to sell. Even the ones that are not tend to cite a single study and call it "evidence-based". The question that is hard to answer — how much research exists, and what kind — has a factual answer, and it is the same answer for everyone.
Where the data comes from
Studies: PubMed
Every study on the site is a real PubMed record, linked so you can read the abstract yourself. Studies are fetched per substance and per health goal, then filtered in code: the substance must be named in the title or abstract, animal and in-vitro work is excluded, and trial protocols (no results yet) are dropped.
Confidence Level: a formula, not an opinion
The signal bars come from counting meta-analyses, systematic reviews and randomised trials that pass that filter. The thresholds are published and applied identically to every substance — no manual overrides. A substance cannot be promoted because it is popular.
Effectiveness Score: editorial
The 0–10 number next to each goal answers a different question: how effective is this substance for that specific goal, according to scientific consensus. It is not derived from study counts — it is an AI-generated editorial rating, reviewed against meta-analyses, and it is the least certain data on the site.
Myths
Each myth page cites the PubMed sources it rests on. Myths without a source that actually addresses the claim are removed rather than kept with a decorative citation.
Where AI is used, and where it is not
Being straight about this matters more than looking impressive. Study extraction is deterministic: publication types, labelled abstract sections and keyword checks, no language model deciding what a paper found. AI is used for the descriptive text — substance summaries, myth explanations, the per-goal Effectiveness Scores and the Ukrainian translations. AI-written text can be wrong in ways that read fluently, which is why every claim links back to a source you can check, and why the report-an-error address below is a real one.
What this site is not
- Not medical advice. It reports what studies found; it never tells you what to take.
- No ads, no affiliate links, no supplement sales, no sponsored placements. Nothing on the site changes if you buy something.
- No accounts, no signup, no tracking of what you search for.
- Not a replacement for talking to a clinician, especially if you take prescription medication.
Found something wrong?
Errors in this kind of data are inevitable: a study counted for the wrong goal, an effectiveness score that does not match the literature, a myth verdict that is too strong. Corrections are welcome and get applied.
Write to [email protected]