No invented people
A fabricated byline is the cheapest way to make an unreviewed healthcare article look reviewed,
and it works on the reader precisely because a name is what the reader has instead of
expertise. Every identifier in authorIds and reviewerIds resolves to a real person who did
the work attributed to them. No composite drawn from several contributors, no pen name standing
in for a team, no plausible-sounding specialist assembled to fill a reviewer slot, and no
credential the person does not hold.
This is a hard rule, not an aspiration. There is no risk level low enough to make an invented
name acceptable, no deadline that licenses one, and no meaningful distance between inventing a
person and inflating a real one — a title never held, a body never joined, a speciality
adjacent to the actual one. A fabricated reviewer is a forged signature. An article found
to carry one moves to WITHDRAWN rather than to a correction: the correction file records what
an article said, and this is a question of who said it.
The checks reach part of this: they confirm that every id resolves to a person record carrying its required fields. Nothing in a static build can confirm that the person exists, holds the credential, or agreed to be named. That control is entirely human, and saying so here keeps a passing build from being mistaken for a verified one.
Where the credential can be checked
Each person record states, for every credential it claims, where that credential can be checked: the register or professional body, the identifier the person appears under there, and the route by which it is read. Where no public register exists for a credential — many do not — the record states how the credential was verified instead: what was seen, by whom, on what date, and how it reached them. The second form is weaker than the first, and it is written as the weaker thing rather than phrased to sound like a register lookup.
This mirrors the rule that a source record requires both dateRetrieved and retrievalRoute:
a register that exists is not evidence that anybody looked in it.
The marketplace's narrow reading of verification applies here without widening. VERIFIED is
not permission to practise and not permission to book; it is a statement that somebody checked
a document, published as a summary of what was checked and never as an endorsement. A person
record here is the same kind of object: it records a check and says nothing about clinical
quality. It is not offered as anybody's standing, and whether publishing a check nonetheless
makes the publisher an endorser is an open legal question marked COUNSEL_REQUIRED, which this
policy does not settle.
Review scope is per subject
A policy keyed on a role can be satisfied by obtaining the role. "Reviewed by a doctor" is satisfied by any doctor at all, including one who knows nothing about the subject and therefore cannot tell whether the article's sources support its claims.
Review scope is recorded per subject, and an article's reviewerIds name a person whose
recorded scope covers this subject rather than one whose general standing is impressive. A
qualification in one field is not a qualification in another, and adjacency is not coverage.
CLINICAL_REVIEW cleared from outside the subject produces an unreviewed article carrying a
reviewer's name, which is worse than an article with no reviewer, because the name is what
stops the next reader asking.
LEGAL_REVIEW is bounded the same way, by jurisdiction instead of clinical field. A reviewer
qualified in one jurisdiction has no scope over another. The United States is never one
corridor — it resolves state by state — so no reviewer's scope covers "the United States", and
the content model refuses a bare US in a claim's jurisdiction for the same reason.
The team byline
GMedAC editorial exists as an author for institutional pages that are genuinely
collective — policy documents like this one, index pages, correction listings — where naming
one individual would misattribute the work of several. It never appears in reviewerIds on
clinical content, and it clears neither CLINICAL_REVIEW nor LEGAL_REVIEW.
A review is an attestation, and an attestation needs somebody who can be asked what they read, be wrong about it in public, and be removed from the record when they were. A team is none of those things. A collective cannot attest. The team byline is also not a way to give a clinical article the appearance of authorship when no individual will put a name to it. If nobody will, that is the article's answer.
A marketplace listing is not a qualification to review
A professional listed on the GMedAC marketplace has been through a document check and holds a
commercial interest in demand for their own field. The marketplace is at Phase 0 — no
checkout, no live payment provider and no deployment — so this rule is written before the
first listing exists rather than after a case. Neither fact says anything about whether
that person can judge whether an article's claims are supported by its sourceIds. Treating a
listing as a reviewing credential would promote a marketplace onboarding step into an editorial
one and put the journal's review capacity on the commercial side of the company. A listed
professional may review, on subjects their recorded scope covers, with the listing declared in
their person record and carried into the article's conflicts declaration. The listing is never
the reason they were asked.
What a named reviewer is attesting to
A reviewer named on an article has read that article, in the revision that is published. The
name attests that they read it, checked its claims against the sources it cites, and consider
those claims supported at the article's contentRisk level. It does not attest that they agree
with every editorial choice, or that they wrote any of it.
No automated check can prove that a reviewer read anything. The build confirms that every id in
reviewerIds resolves to a reviewer record; that an article whose risk and type require
clinical review names a reviewer carrying that article's section in their recorded scope; and
that a jurisdiction-specific claim is covered by a reviewer's recorded jurisdictions. A section
is coarser than a subject, nothing refuses an article whose author reviewed their own work, and
no check distinguishes a careful reading from a scroll. The journal has no database, no
authentication and no analytics, so there is no session, no timestamp and no trail to consult —
and building one would produce a record of attention rather than of understanding. The
attestation therefore rests on a person's willingness to have their own name on the page. The
name is the control, and the person is named for exactly that reason.
Two things follow. A reviewer's name stays attached to the revision they actually read, so a
material change requires review again — a material: true correction with no reviewer is
refused, as is one with dateModifiedChanged: false. And a reviewer may withdraw from an
article at any time without giving a reason, on which an editor returns the article to
CLINICAL_REVIEW or LEGAL_REVIEW. Honouring a withdrawal is a human commitment; only its
consequence is enforced, because a status other than PUBLISHED or UPDATED reaches no
page, no sitemap, no feed and no search index. Making withdrawal costly would make
staying attached the easier choice, which is the opposite of what a name is for. This is
honest, not clean.