How we work
This platform runs on four principles. They are not marketing words — they are technically and editorially enforced.
Aggregation instead of scraping
We pull raw data from open APIs — PubMed, Europe PMC, PubChem, ChEMBL, regulatory publications (FDA, EMA, TGA). We do not copy third-party text. Every summary is written in our own words, our own synthesis. This protects against copyright issues and against inheriting third-party inaccuracies.
Evidence tier visible
Every claim carries its visible evidence tier: human RCT, human trial, animal model, in-vitro, preclinical, anecdotal, theoretical. We do not use weaker tiers as recommendation — they are flagged as weak evidence.
“Observed” ≠ “established”
Every study observation comes with what does NOT follow. Animal model is not human. In-vitro is not in-vivo. An RCT with 12 people is not evidence of efficacy in the general population.
Compliance gate as build check
Before every build and every server start, an automated guard scans content against a list of forbidden phrases: dosing language, usage or hygiene instructions, purchase calls, healing promises, stack protocols. Forbidden field names like `dosage`, `protocol`, `buyLink` also block publishing. If they're not clean, nothing goes live.
- No dosing or usage instructions for humans.
- No healing promises or health claims under the EU Health Claims Regulation.
- No purchase recommendations, no supplier hints, no vendor lists.
- No user quotes as recommendation — anecdotes are marked as weakest evidence.
- No tools that interpret individual values or recommend actions.
Studies count as facts if they come from PubMed, Europe PMC, an indexed peer-reviewed journal with DOI, or a regulatory publication. Forums, vendor sites and influencer content NEVER enter our factual base — they can only ever be marked as "anecdotal" with explicit caveat.