The dietary supplement category is the first consumer vertical where a generation of shoppers is asking an AI engine — not a search engine, not a forum, not a friend — for product recommendations. When a parent types "best magnesium for a kid who can't fall asleep" into ChatGPT, the engine returns named brands, mechanism claims, and dosage guidance in the same paragraph. The brand named has not paid for placement. The claim attributed to it may or may not be a claim that brand itself ever made. Nobody is auditing what the engines say.
We started EvidenceSignal because we couldn't find anyone doing the work seriously. The well-funded AI-citation companies measure citation share but ignore regulatory context; the regulatory-compliance vendors monitor your own copy but ignore what AI says about you. Both halves of the problem matter — together, even more.
What we do
We measure how the four major AI engines — ChatGPT, Perplexity, Claude, and Gemini — recommend supplement brands. We classify every claim those engines make against the largest indexed corpus of federal and self-regulatory enforcement in the category. We publish a daily Index of the supplement AI shelf. We ship a free scanner that gives any brand a one-page scorecard in under a minute. We sell continuous monitoring and pre-publish compliance tooling.
What we believe
One — AI is the new shelf. A supplement brand's most valuable surface is no longer its product page. It is whatever ChatGPT says about it when a real person asks a real question. Brands that don't measure this are flying blind.
Two — compliance and growth are the same problem. Disease claims that trigger NAD challenges also trigger AI hallucinations. Cleaning up your copy for one regulator is the same work as winning your AI shelf.
Three — the corpus is the moat. We cannot be undercut on the work of reading and classifying every Warning Letter back to 2017. That work is six months in already and it compounds weekly.
What we're not
We are not a law firm. We do not give legal advice. We are not an SEO company. We do not promise to "rank you higher" on AI engines through technical fixes. We are not a content agency. We will tell you when your claims are over the line; we will not rewrite them for you.
How we work
Open methodology. Every chart on our site shows its source and date. Every score we publish can be replicated. See methodology.
Honest naming. We name brands in our research when warranted. We are not trying to be everyone's future customer at the cost of saying what's true now.
Institutional voice. Our research is published under the EvidenceSignal Research byline. The work stands on its own.