pepmg_

● Evidence methodology

Trust the receipt, not the confidence.

pepmg’s research layer is built so unsupported citations, figures, evidence labels, and dosing language fail before publication. The checks below run as part of the release process.

Published reviews

70

Distinct cited PMIDs

486

Ranking rule

Affiliate-blind

● The build-blocking checks

What must be true before a review ships.

01

Every citation must resolve

A review cannot ship with a PMID that is absent from the committed source corpus. Citation numbers in the prose must map to the reference list in both directions.

tests/research-seeds.test.ts
02

Numbers must trace to a source

Percentages and decimal results in review prose are checked against the abstracts of the sources that review cites. Unsupported figures fail the build.

tests/research-numbers.test.ts
03

Evidence labels cannot outrun the studies

A review cannot receive a stronger label than its cited study types and species support. Strong and moderate labels require qualifying human evidence.

src/lib/research.ts · tests/research.test.ts
04

Study summaries cannot invent findings

Machine-written study summaries are token-checked against the source abstract. New numbers, verdict language, and unsupported magnitudes are rejected.

src/lib/studySummaryGate.ts · tests/study-summary-gate.test.ts
05

No public dosing instructions

Research content is checked for second-person and imperative dosing language. pepmg describes published research; it does not prescribe a protocol.

tests/research-seeds.test.ts
06

Failure means no deploy

These are executable checks in the normal release suite, not editorial promises. A failed integrity test stops the same pipeline that publishes the site.

package.json · scripts/ship.sh

● What this does not prove

Correct sourcing is not clinical certainty.

These checks prove that pepmg’s summaries stay tied to the sources they cite. They do not prove that a study was well designed, that its result will replicate, that a vendor product matches a paper, or that any finding applies to personal use. Evidence labels describe the published record; they are not recommendations.

Browse the evidence library → · Read the price methodology → · See how pepmg is funded →