Public scientific standard
This is the operational contract behind the research pages: what can raise confidence, what cannot, how conflicting evidence is handled, what automation is allowed to do, and which commercial incentives are forbidden from affecting scientific conclusions.
Canonical grading rules
Grades describe confidence in a defined claim or profile conclusion. A strong grade cannot be created by citation volume, mechanistic plausibility, popularity, or commercial value alone.
Strong
Requires a mature, direct human evidence base: replicated relevant human outcomes or a high-quality synthesis of relevant human trials, acceptable consistency, and no major directness problem that would make the public claim broader than the evidence.
Moderate
Meaningful direct human evidence exists, but replication, precision, duration, population breadth, consistency, formulation match, or study quality still limits confidence.
Limited
Human evidence is sparse, small, indirect, inconsistent, exploratory, or strongly dependent on a narrow preparation/population. Mechanistic or preclinical evidence may add plausibility but cannot promote a claim to B by itself.
Preliminary
Evidence is mainly mechanistic, animal, in-vitro, traditional, uncontrolled, or otherwise too indirect to support a practical efficacy conclusion. The page may remain useful as a research reference.
Do not infer benefit
Evidence is absent, materially unfavorable, too unreliable to support the claim, or the practical framing is limited by a safety context that makes casual recommendation inappropriate. Safety certainty and efficacy certainty remain separate fields.
Inclusion, interpretation & safety
Include sources that actually test or materially inform the ingredient, preparation, outcome, population, safety question, interaction, or mechanism being discussed. Primary human studies are preferred for efficacy claims; high-quality systematic reviews and meta-analyses can summarize a body of evidence.
Do not count a paper toward a claim when it studies a different ingredient without a defensible bridge, only mentions the ingredient in background text, is duplicate reporting of the same study population, or cannot be identified well enough to audit. Retracted or invalidated work should not support a current conclusion.
Do not average disagreement into a falsely smooth conclusion. Separate supporting, mixed, null, and contradicting evidence, then inspect study quality, preparation, dose, population, duration, outcome definition, and risk of bias. Strong wording is reduced when material disagreement remains unresolved.
Animal, cell, receptor, pathway, and biochemical evidence can explain plausibility or generate hypotheses. It does not establish a human benefit, an effective consumer dose, or clinical equivalence. Mechanism and outcome labels remain visibly separate.
Evidence from a specific branded extract, standardized preparation, plant part, salt, chelate, delivery system, or formulation is labeled as such. Findings are not generalized to every product sharing the ingredient name unless the evidence supports that generalization.
Safety is evaluated independently from efficacy. Contraindications, adverse events, pregnancy/lactation concerns, organ-specific cautions, dependence potential, dose-related toxicity, and high-risk populations require source-backed treatment appropriate to the seriousness of the risk.
Documented clinical interactions, pharmacokinetic interactions, pharmacodynamic additive risks, case evidence, and theoretical mechanisms are not presented as equivalent. Missing interaction data means “no documented interaction in our dataset,” not “safe.”
Studied doses are reported as research context, not automatically as medical dosing advice. Preparation, extract standardization, elemental amount, route, population, and duration must match the statement. “No standardized medical dose” is distinct from “doses studied in trials.”
Commercial eligibility comes after evidence and safety evaluation. Product-quality criteria may consider formulation match, label clarity, clinically studied form, third-party testing, and reliability, but affiliate payout cannot change an evidence grade, safety flag, or scientific conclusion.
AI & editorial automation
External review
A page may display an external reviewer’s identity, qualifications, and review date only when a real, page-specific review event has been recorded. The site does not imply physician, pharmacist, dietitian, or other professional review where it did not occur.
Funding & independence
Supplement companies cannot pay to raise or alter an evidence grade. Affiliate commissions cannot alter evidence grades, safety conclusions, source selection, or the scientific inclusion threshold. Product-quality scoring, when used, remains separate from efficacy evidence.
Why trust this page?
The useful question is not whether a page sounds confident; it is whether its claims can be traced to identifiable evidence, whether limitations remain visible, whether review dates represent actual review, and whether material corrections remain public. Those are the constraints this policy is designed to enforce.
Tool context
Exact evidence grading, source inclusion, conflict resolution, branded-extract, safety, dosing, affiliate, AI automation, review, and correction… Tool and utility pages are meant to support research decisions, not replace the full evidence review. Use the results as a triage layer that points you toward the profile, guide, or safety topic that deserves closer reading.
For Editorial, evidence & automation policy, the most useful approach is conservative: check one question at a time, compare the result against the full profile, and avoid assuming that an automated output proves safety, effectiveness, or correct dosing.
If the question involves prescription medications, pregnancy, breastfeeding, chronic disease, surgery, addiction risk, or a child, the next step should include a clinician or pharmacist rather than relying only on a supplement website.