How we use AI

We use AI to source and verify the clinical studies, and to draft text. A human reviews everything before it publishes.

The literature work happens first, before any text exists. AI searches the published research on a question, pulls the studies that address it, and extracts what each one found: the population, the numbers, the effect size, the limitations. It also runs the mechanical checks that are easy to get wrong at volume. Does this reference resolve to a real paper. Do the authors and the year match. Has it been retracted since publication. Does the paper contain the finding attributed to it. Machines outperform people on this work because it is repetitive, high volume and unforgiving of fatigue. It also runs again later. Pages are rechecked against new research on a schedule, and the same verification reruns on citations that passed the first time.

AI writes the first draft from the sources gathered in step one, not from its own memory. A language model asked to write about scars from memory produces fluent, plausible, occasionally invented content, including references to papers that do not exist. Given a set of verified sources and a rule that every claim must map to one of them, it produces something checkable. The format is also constrained tightly: a one-line answer, then facts, then graded treatments, then sources. Rigid structure leaves less room for confident vagueness. Drafts still come back with problems. A number drifts outside its range, a finding gets applied to the wrong scar type, a hedge goes missing. The next step catches those.

Nothing publishes without a person going through it, and the human step is a full check, not a skim. It confirms the draft says what the sources say: numbers back against the papers, claims back against the studies, grades back against the strength of the evidence, and every citation opened. Anything untraceable gets cut. The human step also handles what a model does badly, which is deciding what to leave out. Much of the judgment in health writing is omission: one striking figure from one small study does not belong on a page people will act on. When the check keeps finding the same kind of error, that pattern goes back into the instructions for the next draft. That feedback loop does a large share of the work.

The human step in our workflow is an editorial check, not a clinical consultation. No doctor has examined this page and signed it off, and no page here can be applied to your scar, because we cannot see it. This is an independent reference site, and what it can do is show its work. Every claim points to a source you can open and judge, and every treatment carries a grade indicating how much weight the evidence takes. Use it to understand your options and to arrive at an appointment with better questions. If a scar is painful, growing, tightening across a joint, or a wound is still open past a few weeks, see a clinician.