Free checklist for anyone writing health information
Last month I built something I have no intention of charging for and it is worth explaining why.
Going through PIF TICK assessment changed how I write. Not because the criteria surprised me, but because somebody finally showed me the standard. Most clinicians writing for their own websites have never seen it. They are not cutting corners. Nobody has ever told them "what good looks like".
So I took standards that are already free and public, the NHS digital service manual standard for creating health content and the ten published PIF TICK criteria and turned them into an AI checklist you can actually run.
It works both ways round. Paste in a page you have written and it returns what to fix, with the replacement wording supplied. Or use it as the checklist to write against before you start.
Installing it takes about five minutes: download one file and follow the illustrated guide for whichever assistant you use, Claude, ChatGPT or Gemini. There is also a printable one-page version needing no AI at all, which is what I would hand a sceptical colleague.
Let me be clear about what it is not. It is a self-check. It awards nothing, carries no mark and is not endorsed by PIF or NHS England. Certification assesses how you produce information over time, not how one page reads. If you want the TICK, apply for it. What you can honestly say is that your content is written to the NHS standard.
Which brings me to AI. AI-generated slop has received a bad name quite rightly, but used well it can improve our own writing: it can flag medical jargon, the confusing risk profile numbers, the risk section that quietly became a benefits section and lots more. Used badly it is dangerous, because it will cheerfully invent a plausible risk figure if you let it. So the checklist's hardest rule is that it must never supply a number and it will tell you to delete any figure without a source.
It also refuses to tick certain things: whether a clinician checked it, whether a real patient understood it and whether anyone reads the feedback. All AI needs a "human in the loop" and this is no different.
That refusal is the point. AI can clear the ground. It cannot tell you your reader understood you. Only your reader can.
Free to use and adapt: github.com/alistairphillips1/health-content-check