Two useful components are not a workflow. Here are the connected layers between a prompt and a publish-ready, multilingual medical asset.
“ChatGPT plus a translator” gives you a draft and a language version. It does not give you a workflow. Between a prompt and a page a patient can trust sit a dozen connected steps — topic choice, clinical input, patient-journey placement, terminology control, compliance, localisation, search structure, images, metadata, clean HTML, structured data, social reuse and a final approval. Skip them and you have text, not a publish-ready asset. The point is not that ChatGPT is useless — it is one component of many.
The tempting math: ChatGPT writes it, a translator converts it, done. Then the doctor spends an evening fixing claims, the page has no metadata or images, the translation reads slightly off, and it still isn’t really published. The two tools did their bit; the workflow was missing.
The missing layers
topic validation · clinical input · patient-journey placement · terminology control · compliance (per market) · localisation, not translation · SEO structure · images + alt text · metadata · clean HTML · structured data · social reuse · final clinical approval.
Each is small; together they are the workflow.
Traced through one page
Take a breast-surgery clinic in Budapest wanting a German-language page for patients travelling from Austria and Germany. “ChatGPT + translate” produces a fluent English draft and a fluent German version. What is still missing tells the story:
- Topic and intent — does the page target what German-speaking patients actually search, or a translated English keyword that no one uses?
- Clinical input — does the pathway match how this clinic operates, and are outcomes framed honestly?
- Compliance — Germany’s HWG is strict on aesthetic advertising; a literal translation of English marketing wording can breach it.
- Terminology and protected terms — device and technique names must be preserved exactly, not “translated”.
- The assets — images with alt text, metadata, structured data, a clean publish format — none of which the two tools produced.
The draft and the translation were the easy 20%. The workflow is the 80% that decides whether the page is trusted, found, and safe in its market. 〈CLAIM: illustrative, HWG example is directional〉
Where AI helps, and where it needs control
AI does real work inside this chain — drafting, structuring, simplifying, first-pass translation. What it does not do is validate the topic against your services, enforce national advertising rules, localise for how each market actually searches, or sign off the medicine. It is a powerful component, not the whole line.
Common questions
- Can’t I just add the missing bits myself?
- You can — that manual glue is the hidden cost, and it usually lands on the doctor’s evenings.
- Isn’t machine translation enough for the languages?
- For gist, yes; for medical meaning, terminology and market compliance, no — that is localisation, not translation.
- So is AI worth using at all?
- Absolutely — as components inside a controlled workflow, not as the whole of it.
Turn one treatment into a complete patient-education content block
The whole workflow, produced and publish-ready — not just a draft.
