How a page is approved before it is published

Most sites ask you to trust a check you cannot see: a photograph, a title, a string of letters after a name, and the suggestion that somebody qualified read the page before you did. What follows does the opposite. It describes the check itself — the stages a text passes through here, the role that signs it, what a signature is made of, and what happens to a page that nobody signs. Two names sit at the foot of the content pages and nowhere else: no credential class, no registration number, no profile, no link out. Our editorial standards give the reason for that silence — we hold no document that would support a credential, and printing nothing beats printing a qualification we cannot evidence. What is published instead is the mechanism, in enough detail that you can ask whether it ran.

What the approval line actually is

Approval here runs in four stages and a page moves through them in order. Sourcing and wording come first, and our methodology sets out that route for a single claim in detail. Then every paragraph of a content page is labelled, so the page carries a record of what kind of statement each passage is and which of them carry clinical judgement. Then the page is read by the medical editor, who is not the person who wrote it. Then, and only then, a signature is either written or withheld. Nothing in that order is left to discretion, and a page does not skip a stage by being well written. What has not reached the last stage stays a draft.

What a signature is made of

A signature on this site is not a name printed under a page. Each record holds the page it belongs to, the language, the approving role, the date, and a hash computed from the text of every passage on that page that carries a clinical proposition. The hash is what makes the signature perishable: rewrite one of those passages and it no longer matches, the signature stops being valid, and the review date disappears rather than standing over text nobody approved. Our editorial standards say a reader can see who checked a page but not when, and the hash is the part that would change that, because a review date produced any other way cannot reach a published page at all.

What happens when nobody signs

Publication is withheld, and that is the whole of it. A page that carries a clinical proposition and holds no signature does not appear at all. Counted on 10 September 2026, 41 of the pages written for this site carry at least one passage classed as a clinical proposition, and all 41 have a recorded signature; our methodology explains that a single clinical proposition is enough to place a page in that group. The dates are 3 September 2026 for the first two, 5 September 2026 for the next three, 6 September 2026 for the next twenty, and 10 September 2026 for the sixteen Arabic pages. One approver is on record, the medical editor, and he signed all of them: 11 of the 11 English clinical pages, 14 of the 14 French ones and 16 of the 16 Arabic ones. That is a concentration of 100 percent in each of those three lines, and we print the figure rather than leave it to be worked out. Our ceiling is not a fixed share but one that tightens as approvers are added — it stands at 100 percent while a single person signs, halves at two, and settles at 25 percent from the fourth onward, where it stays. With one approver it binds on nobody, and printing the concentration is what stands in its place.

Where this line is thin

Three weaknesses belong on this page rather than in a reader's discovery. The approving role is this site's own medical editor and not an outside clinician, so the check is internal; no clinician from outside this site has read any page here, and this line does not pretend otherwise. No registration number is published for anyone in the approval record, because none is held, and an empty field is evidence of nothing. And the ceiling described above was applied by hand when those signatures were written: the check that tests it now runs against the real signature record every time this site is built, where until 6 September 2026 it had only ever run against synthetic examples. Our corrections log is where any of this changes on the record.