A study can be interesting without being a reliable answer. Our human evidence ratings describe how much confidence we place in a finding for a specific supplement, use, and group of people. They are designed to make the uncertainty easier to see.
How certain is the evidence? And what does it say about benefit or harm? A high rating does not automatically mean a supplement works.
Yes. GRADE is an established approach for assessing certainty across a body of research for a particular outcome. It uses four levels: high, moderate, low, and very low. It considers study limitations, differences between findings, relevance to the question, uncertainty in effect estimates, and possible missing research.
Root & Evidence uses its own five-label editorial framework, described below. The first four names are familiar from GRADE; the fifth, “No eligible human evidence,” describes the absence of qualifying research and is not a GRADE level. Our ratings are not formal GRADE assessments, a validated scoring instrument, or a professional certification. We will only describe a review as using GRADE if it follows that framework’s requirements and publishes the supporting assessment.
Each rating belongs to a defined question, such as a particular magnesium formulation for insomnia symptoms in adults. A supplement can receive different ratings for different uses. These descriptions are decision rules for editorial judgment, not an automatic point system.
A sufficiently complete body of directly relevant human research supports a precise, consistent conclusion, with no important unresolved concerns about bias or missing results. Appropriately designed randomized evidence generally anchors conclusions about treatment effects. Further research is unlikely to materially change the conclusion.
Relevant human evidence supports a conclusion, but an important limitation leaves room for it to change. For example, estimates may retain meaningful uncertainty or applicability may be somewhat limited. The published rationale must explain why those concerns reduce confidence without making the finding highly uncertain.
Human research exists, but serious limitations make a different result plausible. Examples include substantial risk of bias, inconsistent findings, indirect populations or formulations, or estimates too imprecise to distinguish meaningful benefit from little effect. Better research could materially change the conclusion.
The available human findings are too fragile to support a dependable conclusion. This may reflect very serious flaws, several important limitations together, or uncontrolled observations that cannot separate the supplement’s effect from other explanations. A positive result at this level remains highly uncertain.
A documented search has not identified an eligible human study for the exact question and search date. Cell experiments, animal studies, traditional use, and biological plausibility can explain interest, but do not fill this gap. This label means evidence is absent for the defined question, not that the supplement has been proved ineffective.
Our current research guides summarize selected trials. They have not undergone the complete search and appraisal required for these ratings, so they display “Not yet rated.” We will not assign “No eligible human evidence” simply because we haven’t completed a review.
Before assigning a rating, we require a written review record that another reader can follow. Study count alone does not determine confidence. One very large, rigorous trial may be more informative than several small, unreliable studies, while a meta-analysis can still inherit the weaknesses of its included studies.
Beside the rating, a completed review should describe the result as benefit supported, little or no meaningful effect, uncertain or inconsistent results, or harm suggested. It should also report the size of the effect and relevant limitations. “Not statistically significant” alone does not prove that there is no useful effect.
High certainty + little or no benefit: rigorous research can give us confidence that a supplement adds little for a particular outcome.
Low certainty + apparent benefit: an encouraging finding may still be too uncertain to rely on.
A human evidence rating is not a safety clearance, a product-quality score, or a recommendation for your personal situation. It does not establish that a product on a store shelf matches the preparation studied. Adverse effects, interactions, pregnancy, medical conditions, and product testing need their own discussion.
Neither popularity, an Ayurvedic or Traditional Chinese Medicine label, an affiliate relationship, nor a convincing biological story raises a rating. Traditional use can provide context and research questions; evidence about outcomes in people is assessed on its own merits.
Methodology published November 3, 2026. This page defines our editorial policy; it does not claim endorsement by GRADE or Cochrane.