Medical review service
Your team is publishing AI-generated
medical content. Someone should be
reading it as a doctor.
Clinical review of AI-assisted healthcare content. Specific findings, severity ratings, and rewritten copy you can publish.
Where it actually fails
Two sentences. One word apart.
Your writer will not see a difference. Neither will your editor, or the model that wrote it.
Defensible
“Our algorithm flags suspicious lesions for radiologist review.”
Matches what imaging AI is typically cleared to do: mark something for a qualified human to assess.
A claim you may not be able to support
“Our algorithm detects suspicious lesions for radiologist review.”
Describes the software finding something. That is a statement about intended use, and it may exceed your clearance.
Detects. Diagnoses. Treats. Prevents. Causes. Proven. Each one moves a sentence from something you can defend to something you cannot.
What it costs when it goes wrong
Regulatory exposure
Marketing copy is discoverable. A claim about intended use that exceeds your clearance is a problem whether it appeared in a submission or a blog post.
Trust, with the audience you need most
If your buyer is a clinical department, clinicians read your marketing. One overstated claim tells a specialist nobody clinical reviewed the page.
Patient safety
In patient-facing content, advice that is subtly wrong about dosing, timing or when to seek help causes real harm, and it surfaces long after publication.
A very bad fortnight, eventually
A two-year-old page surfaces during a funding round or procurement review, and someone audits eighty pages at once under time pressure.
What you get
A findings table
Every issue with a severity rating and a recommended action. Critical items flagged separately so you know what to pull today.
The clinical reasoning
Not just that something is inaccurate, but why, what the correct position is, and what it exposes you to. Your team learns the pattern.
Copy you can publish
The problem sections come back rewritten. This is the part most reviews skip, and the part that saves your team a week.
A claims guide
Recurring errors written into a short reference your writers work from, so the same mistakes stop being made.
How it works
01
Send the content
A single page, a batch, or a sitemap. Tell me your regulatory position and who the content is for.
02
Triage by risk
I rank by exposure, so the highest-risk pages get attention first rather than working alphabetically.
03
Review and rewrite
Findings documented, problem sections rewritten and ready to publish.
04
Debrief
A short call on the patterns, and what to change in the brief so the same issues stop arriving.
Engagement options
Single page review
One high-stakes page. A landing page, a product page, or a claim you are unsure about.
Content audit
A batch of existing pages, triaged by risk. The best first step if nobody has looked in a while.
Pre-publication review
New content gets a clinical read before it goes live. The cheapest point to catch anything.
Ongoing partnership
Monthly retainer covering pre-publication review plus rolling work through the back catalogue.
See an actual review
I have published a full worked example.
A clinical audit of an article headlined “How AI Diagnoses Cancer With 99% Accuracy.” It includes the findings table, the reasoning on each issue, and three full before-and-after rewrites. It is the exact deliverable you would receive.
Read it before you decide whether this is worth your money.
Questions
- Are you against using AI for content?
- No. I use it myself for parts of the process. The argument is not that AI-assisted content is bad, it is that fluent output creates false confidence, and in healthcare the cost of a confident error is higher than in most categories.
- Will you tell us not to publish things?
- Sometimes, and I will always explain why and offer a version you can publish instead. My job is to find you the strongest claim you can actually defend, not the safest one.
- What if our content was written entirely by humans?
- Same service, same process. The failure modes overlap more than people expect.
- How fast is it?
- A single page is usually a same-week turnaround. Larger audits depend on volume, and I will give you a date before starting.
- Do you sign NDAs?
- Yes.
Find out what is currently live on your site.
Send me one page you are unsure about. I will tell you what I would flag and why, and you can decide from there whether a full review is worth it.