Skip to content
GMedACJournalGMedACJournal
SearchGMedAC

Medical content policy

What clinical content this publication carries, what it refuses by name, and which refusals the build can enforce

A worried reader will act on a sentence that sounds as though it was written for them. This publication is written for populations, never for a person. It does not provide personalised medical advice, it has no way to provide it, and nothing here — no article, no worked example, no comparison table — should be read as an assessment of the reader.

What is never published

Each item below is refused by name so that a later editor cannot argue it into a draft as an edge case.

  • Patient-specific diagnosis. No article tells a reader what condition they have, and none builds a decision path that arrives there by a longer route.
  • Treatment plans for an individual. Describing how a condition is generally managed is reporting. Telling a reader what to do about their own case is not reporting, whatever hedging surrounds it.
  • Medication dosing. No article carries a dose, a schedule, a titration or a conversion, even where a public guideline states one — reproducing the number strips away the assessment that made it appropriate for somebody.
  • Instructions to begin or discontinue a medicine. That decision belongs to the prescriber who can see the patient and the record. This publication can see neither.
  • Claims of a guaranteed outcome. No treatment is described as certain and no procedure is described as free of risk.
  • Emergency triage presented as a medical assessment. The publication cannot judge urgency and refuses to imply that it can.
  • Unsupported safety or effectiveness claims — any statement about benefit or harm not carried by a sourceIds entry strong enough for the article's declared risk level.
  • Individual clinician recommendations dressed as journalism. Commission does not influence organic ranking in the marketplace and a test pins that; an editorial page naming favoured professionals would route around the rule rather than respect it. The journal carries no advertising and no sponsorship today, and any future advertising surface would be a decision taken openly, not an editorial habit that grew.

A fixed list of dangerous phrasings fails the build outright when it appears in prose. That check is real and it is shallow: it matches wording, not meaning. A paragraph that reaches the same conclusion in careful language passes it untouched. The phrase list buys nothing except the elimination of the crudest mistakes.

Why no reader can be assessed here

The site is static. output: export makes the framework refuse cookies(), headers(), searchParams and middleware at build time, there is no database and no authentication, and search runs entirely in the reader's browser against a static JSON index. Nothing about the person reading is available to anything that renders, so no page can be aimed at a particular reader even by accident. That forecloses targeting and nothing else: a paragraph written in the second person still reaches somebody who takes it as written for them, and no check in the build can tell the difference. Keeping the writing addressed to a population is an editorial commitment, not an architectural one.

The consultation happens outside GMedAC. The professional decides whether to accept a patient, what the consultation contains, and everything clinical that follows from it. GMedAC holds no clinical record and issues no clinical document. An article that told a reader what a clinician should conclude would be narrowing that independence from a page the clinician never saw.

Emergencies

An article a reader might reach while something is going wrong declares emergencyInformation: true. Such an article points to local emergency services and states that this publication cannot assess urgency — a requirement on the editor, because nothing reads the body to confirm that either sentence is there. What the flag does mechanically is narrower and certain: a stale article carrying it fails the build at any risk level, because an out-of-date emergency page is worse than no page at all.

Named review for HIGH and CRITICAL

An article of a clinical type — a patient guide, a professional guide, an evidence summary — declaring contentRisk of HIGH or CRITICAL does not publish unless one of its reviewerIds names a reviewer whose recorded clinical scope covers the article's section. A cardiologist's name does not clear a page about pregnancy. Scope, not seniority. The check compares two declared fields, so it stops the reviewer who was never plausible for the subject and not the reviewer who signed off at the edge of what they know.

Any article at HIGH or CRITICAL, clinical type or not, may rest only on sources at or above PEER_REVIEWED_STUDY in the source hierarchy, so INSTITUTIONAL_SUMMARY, PROFESSIONAL_BODY_POSITION, JOURNALISM and COMPANY_STATEMENT cannot carry a high-risk claim. Every source record also carries dateRetrieved and retrievalRoute, the second in prose, because a successful HTTP status is not proof that anybody read the source.

An article held back for want of a qualified reviewer holds at CLINICAL_REVIEW, and that status is not a page: only PUBLISHED and UPDATED reach a page, a sitemap, a feed or the search index. There is no route where an absent reviewer is substituted by a longer warning. Blocked is the correct state, not a delay to be worked around.

A disclaimer does not repair unsafe content

A disclaimer changes what a reader is told about the text. It does not change the text. If a paragraph would mislead a reader into a decision about their own care, a notice above it leaves that paragraph exactly as dangerous as it was and adds a paragraph nobody reads. Disclaimers in this publication exist to set expectations about scope, not to license content that failed review. A disclaimer is not a control.

The risk level is declared, not detected

contentRisk is set by the editor preparing the article. Every mechanism in this document keys off that field: the reviewer requirement, the source floor, the freshness failure. An article about medication interactions filed as LOW receives LOW's rules, publishes without a clinical reviewer, and fails nothing. The build cannot read prose for danger and does not pretend to.

This is the failure mode this policy cannot detect. It is not mitigated by a stricter rule downstream, because everything downstream reads the same declared field. What stands against it is thin: a second person in EDITORIAL_REVIEW, the fact that authors and reviewers are named on the page, and a corrections file that records previousClaim verbatim next to correctedClaim so that a wrong call stays visible after it is fixed. A material correction is refused without a reviewer and refused when dateModifiedChanged is false, which makes quiet repair harder, but none of that catches the article filed at the wrong level in the first place. This is honest, not clean.

Under-declaration is therefore treated as an editorial offence rather than a build failure, which is the only place it can be treated at all. Where an editor is unsure between two levels, the higher one is declared — the cost of that choice is a reviewer's time, and the cost of the other choice is borne by a reader.