How a verdict gets earned.

Three steps, the same every time. No step can be skipped, no step can be paid for.

Step 01

Document

Every ingredient entered by hand against the primary literature. 3,862 so far, each with its trail of citations.

Step 02

Decode

Doses checked against what the trials actually used, forms against what was actually studied, labels against both.

Step 03

Verify

For brands, independent lab results read line by line against the label. Labs test. IngredientMD verifies. If it fails, no badge.

The corpus

Where the intelligence comes from.

3,862INGREDIENTS DOCUMENTEDIngredient chrome, site wide.

47,000+CLINICAL TRIALSThe evidence deep reads cite.

59,000+INTERACTIONS ON RECORDDrug and supplement, the check engine.

525PRODUCTSProduct chrome only.

Sources are cited by name: PubMed / NIH, ClinicalTrials.gov, Semantic Scholar, SUPP.AI and the rest of the 10 primary databases below. None of them endorse IngredientMD or any product, and the citations are the point.

The record path

How a formula becomes a record.

Formula assessed

The formula is read against the trials and the interaction record, dose by dose, form by form. Nothing the label claims is taken on faith.

Lab papers reviewed

Papers on file are read for what they are. A certificate that states specifications is said to state specifications, nothing more. Labs test. IngredientMD verifies.

The gate

What holds up earns a verdict from the closed library and a public record. What does not hold up earns silence, never a warning label.

The verdict

The verdict can't be bought or edited. Brands pay for display and tooling, never the outcome.

The fee

Private until you publish, and the fee pays for review time, never for the outcome.

The division of labor

Labs test. IngredientMD verifies. A lab result is evidence; the read of it against the label is the work.

The limits

Limit stamps say checked, never verified. A regulator publishes a number; the engine checks the label against it and says exactly that.

PUBLIC LIMITS · CHECKED

The verdict library

40 words. Closed.

Every public verdict comes from this list, three registers deep. The registers are claim types, not grades, and none is second place: Clinical sells strength, Precision sells fit, Emerging sells momentum. There is no negative word in the library, what doesn't hold up earns silence, never a warning label. The internal score still sorts the library and decides the register. It never speaks in public.

Clinical register18 words, the strongest evidence tier

Backed by human trialsWorks at the label doseThe studied formFixes a real deficiencyMeta-analysis gradeSafe to take long termReplicated resultsFast to feelWell toleratedProven against placeboLab results match the labelNo hidden blendsClean interaction recordAbsorbs the way it shouldDecades of safe useEarns its label claimsKeeps working long termStrong evidence, strong dose

Evidence strength renders as a filled ring in the ingredient science sections; the fill tracks the research depth, never the ingredient.

PubMed / NIHClinicalTrials.govSemantic ScholarSUPP.AINIH DSLDOpenFDAUSDA FoodDataChEMBLRxNormDose-response records10 primary databasesPubMed / NIHClinicalTrials.govSemantic ScholarSUPP.AINIH DSLDOpenFDAUSDA FoodDataChEMBLRxNormDose-response records10 primary databases

What we do not do

Accept payment for a verdict, a review, or a mention. The display license is the only thing for sale.

Say safe or approved. Labs test, we read the records, your doctor makes the call.

Diagnose or treat. This is information, built carefully, and it is not medical advice.

Publish a number we can't source. The old press stats left for exactly this reason.

The correction path

The engine reads everything and flags what needs a closer look. Corrections land at editorial@ingredientmd.com and get a same-week read.

See a record that earned itA live record, batch 5120130. Every section on it is real.

Open the record →