How we use AI
A plain account of what is automated here, what is not, and why we tell you.
The short version
Health Newspapers articles are drafted with AI assistance from primary sources, under human editorial responsibility. We disclose this on every article rather than burying it in a policy page, because we think readers are entitled to know how what they are reading was made.
What the system does
It monitors primary sources — health agencies, regulators, journals, and preprint servers — and clusters items that report the same underlying event. It drafts articles strictly from the text of those fetched sources, with each claim bound to the specific passage supporting it.
What it is structurally prevented from doing
The drafting model never supplies a citation itself. Source titles, publishers, dates, and links are joined in from records of documents we actually fetched, which means a fabricated citation has nowhere to come from. Claims that cannot be traced to a stored source passage do not survive to publication.
Where humans sit
Mac Giles, our editor, holds editorial responsibility for this publication and every article carries his byline. Anything touching drug safety or dosing, an active outbreak, a named living individual, or pending regulation carries a raised risk tier that marks it for editorial review. Corrections are made by people.
What we do not do
We do not generate fictional bylines or author photographs. We do not claim medical review that did not occur. We do not publish AI-written opinion or first-person commentary. We do not watermark or synthetically brand our imagery.
Why disclose at all
Partly because it is required of publishers informing the public on matters of public interest, and partly for a simpler reason: a health publication that is not straight with readers about its own methods has no standing to report on anyone else’s.