How We Review

The review pipeline

• Sourcing: each study is entered into a source registry before anything is extracted from it.
• AI extraction: an AI model pulls dose ranges, bioavailability, and interaction data from the source text. At this stage it is never used in scoring.
• Human review gate: a credentialed reviewer checks each extracted record against the primary source, corrects errors, and marks it reviewed. Only reviewed records feed the score.
• Scoring parameters: once enough reviewed records exist for a compound, a studied range and Evidence Tier are generated and confirmed by the reviewer.

Evidence Tiers

Tiers are derived mechanically from the studies behind a compound, so they are auditable rather than a matter of opinion. Tier A reflects meta-analyses or multiple independent human RCTs; Tier D reflects in-vitro, animal-only, or a single small human study, for which no dose-adequacy verdict is made.

Where to read more

Our methodology page covers the same pipeline in more depth, including how evidence bounds the Spend Efficiency Index and why the score is independent of affiliate relationships.