In healthcare, people won’t book until they trust the page. When I joined Gaba Healthcare, a telepsychiatry company, in February 2023, a page per condition, trust-focused tests and better search did more for bookings than budget could (the case study has the numbers). That work was slow, and it’s where AI helps. The catch is that healthcare marketing touches two things most industries don’t: patient information, and claims about what a treatment does.
1. Condition pages
At Gaba I built a landing page for each condition so the page matched what the person searched for. I’d keep that structure and speed up the drafting.
In: a clinician’s outline for the condition, the questions people ask, and the approved wording on who the service is for.
AI does: drafts the page in plain language, and a short list of questions for the FAQ.
A person checks: a clinician reads every page before it goes live, and signs it off.
Measure: time from outline to live page, and the booking rate once it’s live.
2. A/B test planning and read-outs
In: search terms, page analytics, and last quarter’s test results.
AI does: suggests test ideas tied to what people searched for, and writes the read-out from the results table.
A person checks: I choose which tests run and check the numbers before the read-out goes to anyone.
Measure: tests run per month, and the share that move the booking rate.
3. Search gaps
In: the search terms report. It has no patient information in it.
AI does: groups the terms by question and shows which ones have no good page yet.
A person checks: a clinician decides which topics are safe and useful to write about.
Measure: organic visitors and qualified calls from the new pages.
4. Reminders and follow-up messages
In: approved templates for each stage.
AI does: drafts variants of the wording.
A person checks: compliance approves the template. After that, sends use the approved text.
Measure: show-up rate and replies.
Patient information goes into a tool only if the tool’s vendor has signed a business associate agreement with you; without one, no patient data goes in. A clinician owns every statement about what a treatment does. I wouldn’t build a symptom checker, an AI that tells people what’s wrong with them, and AI never writes or edits reviews.
I’m not a lawyer, so check these with yours. They’re the ones I’d start from.
Trackers on health pages. In June 2024 a federal court in Texas ruled that part of HHS’s guidance on tracking technologies went too far, for public pages that don’t need a login. HHS later dropped its appeal. Privacy lawsuits over pixels and cookies on health sites didn’t stop, so I’d keep ad and analytics trackers off logged-in and booking pages unless counsel says otherwise.
Health claims. The FTC’s guidance for health products says claims need competent and reliable scientific evidence, usually well-designed human studies, and that implied claims count too.
Testimonials. A dramatic result needs a clear statement of what’s typical. “Results not typical” in small print doesn’t fix it.
Reviews written by AI. The FTC’s rule on fake reviews, announced in August 2024, bans creating or buying them, including AI-generated ones.
Week 1: map the tools that touch patient data, and the trackers on each page.
Week 2: write the approval path, who reads what, and how long they have.
Week 3: run the condition page workflow on one condition.
Week 4: compare the time and the booking rate with the old way.
Hours saved from outline to live page, the booking rate on the new pages, and one safety count: the number of drafts a clinician or compliance reviewer sent back. If that number stays high, the inputs are wrong, and I’d fix those before making anything faster.