How we publish

Source traceability is never collapsed

Every source record carries the label from the original research system, and that is the label the reader sees. Where the system needs an internal status, the mapping is explicit and reversible — never the other way around. A search snippet is never promoted to a verified citation.

Six evidence categories, never conflated

1. Scientific evidence — peer-reviewed research or systematic reviews.
2. Clinical / dermatological consensus — established practice consensus.
3. Consumer experience — reported experience. Not clinical evidence.
4. Social-media popularity — a popularity signal. Says nothing about efficacy.
5. Marketing claim — what a brand says. Never presented as science.
6. Own inference — our editorial reasoning, labelled explicitly.

Claim-level evidence

Significant medical or scientific claims are represented as structured records, not as loose prose. Each claim tracks its supporting sources, the traceability of each source, the evidence category and rating, reviewer notes and its review status. Articles link to claims; claims link to sources. When a claim changes, every article using it is traceable.

Publication gates

Our publishing system blocks a story from going live as evidence-backed content when:

• a sentence reads as a medical or efficacy claim but is not covered by a registered, sourced claim record;
• a claim has no supporting source;
• a claim is supported only by search-snippet sources;
• a claim is supported only by marketing-claim or social-popularity sources;
• every supporting source still has an unassessed evidence rating;
• the article discusses prescription or medically supervised treatments without directing readers to appropriate medical supervision and referral guidance;
• the article implies clinician review without a review record from a qualified reviewer;
• the article rests entirely on weak evidence but carries no visible evidence note.

Gates are enforced server-side. A blocked article explains which rule was hit and exactly how to resolve it. A human reviewer can make an explicit override, but the override is recorded in the audit log with a reason — it is never silent.

Weak evidence is labelled, not buried

Where evidence is limited, incomplete or unavailable, the article says so in a visible evidence note. Where a rating has not been assessed, it stays marked unassessed. We do not infer missing ratings, and we do not upgrade a source's trust status without retrieving the underlying source.

Marketing and science are separated

Product pages carry claims as marketed and an evidence note as separate fields, and they are displayed separately. Ingredient-level evidence does not transfer to product-level efficacy — especially across concentrations, where 20% prescription azelaic acid and 10% cosmetic azelaic acid are treated as different evidentiary propositions.

Attribution is truthful

We do not invent author credentials. Articles are attributed to the editorial desk unless a genuine named contributor is recorded. Clinician review is only ever attributed when a qualified reviewer has actually performed and recorded the review — never as a marketing device.

Commercial relationships are disclosed

Articles containing affiliate links carry a visible disclosure. Editorial evidence ratings are determined independently of any commercial relationship. A product is never ranked as clinically superior because it pays a commission.

Corrections and updates

Articles show publication and update dates. When research corrects an earlier finding, we say so and explain the correction rather than quietly amending the record.