How content is planned

Topics are chosen on three grounds: what readers are searching for and not finding a clear answer to, what a clinician would want a patient to understand before an appointment, and where existing coverage online is thin, contradictory or written to sell something.

Before drafting begins, each page gets a brief that states the question the page must answer, the reader it is written for, the sources that must be consulted, and the specific things the page must not do — for example, name a diagnosis or give a dose.

The standard page structure

Consistency is a reader service. Once you have read one condition page you know where to look on all of them.

  • A direct answer in the first screen — the question the reader arrived with, answered before anything else.
  • Key takeaways — the four or five things that matter if you read nothing else.
  • The body, in a fixed order per content type. Conditions run: what it is, symptoms, risk factors, diagnosis, treatment categories, prevention, warning signs, questions for your clinician.
  • Warning signs, stated explicitly, with the threshold for urgent care rather than a vague instruction to see a doctor.
  • Frequently asked questions, drawn from what people actually ask.
  • Sources, linked directly to the primary material.
  • A review stamp showing the label and the date the page was last updated.

How a page is checked before it publishes

  1. Claim check. Every factual statement is traced to a source. Claims that cannot be traced are cut, not softened.
  2. Source check. Each cited source is opened and read. We confirm the source says what the draft claims it says, and that it has not been superseded.
  3. Safety check. A dedicated pass looking for anything that could be read as diagnosis or dosing, any missing red-flag guidance, and any place where a reader in danger might be reassured wrongly.
  4. Clarity check. Read for a reader with no medical background. Jargon is either removed or defined the first time it appears.
  5. Accessibility check. Heading order, link text that makes sense out of context, alt text, table headers, and colour never used as the only carrier of meaning.

Update cadence

Every page carries the date it was last updated, and that date is only changed when the content actually changed.

  • Condition, medication and guide pages are reviewed at least every 12 months.
  • Pages covering fast-moving areas — screening recommendations, newly approved drugs, AI in medicine — are reviewed at least every 6 months.
  • Any page is updated immediately when a major guideline changes, a safety communication is issued, or a reader reports an error we can confirm.

Where a change alters the substance of the advice, we say what changed rather than quietly swapping the text. See the corrections policy for how we distinguish a correction from a routine update.

Plain language, and how we express uncertainty

We aim for prose a capable reader can follow without a medical background: short sentences, defined terms, active voice, and no euphemism where a clear word exists. Body text is set at a minimum of 17 pixels because most health content on the web is too small to read comfortably.

Saying how confident we are

Health evidence is rarely uniform, and pretending otherwise is the most common failure in health writing. We use a consistent vocabulary:

  • "Strong evidence shows" — multiple well-conducted trials or systematic reviews agree, and major guidelines reflect it.
  • "Evidence suggests" — the direction is reasonably consistent but the trials are smaller, shorter, or largely observational.
  • "Early research indicates" — preliminary, unreplicated, or from small or animal studies. Interesting, not actionable.
  • "Evidence is mixed" — good studies disagree, and we say who disagrees with whom.
  • "We do not know" — used deliberately and often. It is a legitimate answer and readers deserve it.

Where authoritative bodies disagree — screening ages are a recurring example — we present both positions and name the organisations, rather than silently choosing one.

AI-assisted drafting, stated plainly

We use AI tools in producing this site, and we would rather say so directly than bury it.

What AI is used for

  • Producing first drafts and structural outlines from an editorial brief.
  • Rewriting passages for clarity and reading level.
  • Suggesting questions readers commonly ask about a topic.
  • Checking drafts for internal inconsistency and for language that strays towards diagnosis or dosing.

What AI is never used for

  • Sourcing. Citations are never generated by a model and accepted. Every source on this site is located, opened and read by a person. Language models fabricate plausible references, and that failure mode is exactly why this rule exists.
  • Final publication. No page publishes without a human editor having read it in full and taken responsibility for it.
  • Medical judgement. A model does not decide what is safe to tell a reader, where the threshold for urgent care sits, or what a page should refuse to say.

Where responsibility sits

With the human editorial team, entirely and without qualification. "The AI wrote it" is not an explanation we will ever offer for an error on this site. If something here is wrong, a person published it, and a person will correct it.

What we will not write

  • No claims about curing, reversing or preventing disease that the evidence does not support.
  • No fear as a device to hold attention, and no reassurance we cannot justify.
  • No fabricated statistics, no invented experts, no made-up case studies presented as real.
  • No content written to fit a commercial relationship. See the advertising policy.
  • No dosing, and no naming of a diagnosis for an individual reader.