Healthcare SEO at Scale

Challenge

The hospital system had deep clinical expertise and very little of it was reachable through search unless you already knew the name. Its condition and treatment pages had thin page-one visibility, and national health publishers held the direct answers to the questions patients ask first.

Publishing more content would not have fixed it. The program needed a repeatable way to decide which topics deserved a page, which URL should own each one, and how those pages should be built. All of it also had to clear clinical and compliance review.

A note on the figures below. This work is covered by confidentiality, so everything here is a multiple rather than a raw count.

Solution

1. 01

   Demand research

   Find out what people are searching for.

2. 02

   Page ownership

   Decide which URL should answer each question.

3. 03

   Prioritization

   Rank the topics we can win.

4. 04

   Content brief

   Settle the page before anyone writes it.

5. 05

   Clinical review

   Doctors sign off before anything publishes.

6. 06

   Measurement

   Track only the pages the program touched.

The aim was to make the same decision the same way every time, at a volume a hospital review process could keep up with.

Map demand to the page that should own it

I mapped tens of thousands of queries to the page that should answer each one. That showed where topics had no page at all, and where several URLs were quietly competing for the same search.

Prioritize what was winnable

I ranked topics on demand, current visibility, how hard the result would be to take, and what the page was worth to the business. The harder calls were the ones I turned down. Plenty of high-volume topics never got a page, because there was no realistic path to the click and chasing them would have burned review capacity I needed elsewhere.

Give every page a brief

I wrote a brief for every page, hundreds of them across the clinical service lines, fixing the URL, the heading order and the question each section had to answer. Headings went in as the questions people type, with the answer directly underneath. Where one page was being asked to cover two intents, I split the topic so each page had a single job.

All of it then went through clinical and compliance review, which is the real constraint on a program this size. Settling the search decisions inside the brief meant reviewers were checking medical accuracy rather than reopening structure, and the queue kept moving.

Measure only what the program touched

I tracked the wider clinical footprint and the pages we built or rebuilt as two separate things. Every result below is the second kind, which is the narrower number and the one the work can be held to.

Outcome

The snippet growth is the largest multiple here and the one that needs the most context. The site held almost none before the program, so the base it grew from was very small even though the finishing number was not.

The result that mattered most is unbranded clicks. The original problem was that nobody found this hospital system without typing its name, and that is the number that says whether it got fixed.

• All service line pages
• Pages we built or rebuilt

100300500700900

First new page published

StartYear 2Year 3End

The pages we built or rebuilt finished the period well clear of the wider clinical footprint, and were still climbing when the engagement ended.

It was not evenly spread. The strongest service line multiplied its keyword coverage around thirty times over. The weakest barely moved, and those were mostly the ones that published late and never got time to compound.

Not every metric moved the same way

Some page groups lost traffic because I deliberately narrowed what they were for. Average position stayed roughly flat while the ranking footprint expanded underneath it. I read visibility, traffic, conversions and engagement as four separate questions, and a program this size will always have some of them pointing in different directions.

Why it still holds up

This work predates AI Overviews. It was built for organic search and featured snippets, and calling it an AEO program in hindsight would be revisionism.

What carried forward were the fundamentals: clear page ownership, a direct answer near the top, structured information and content organized around real questions. Those still make a page easier for a machine to read. Whether these pages are cited by assistants today hasn’t been measured, so it’s not something I’ll claim.

I owned the keyword framework, the brief system, the page standards and the measurement model, and reviewed the work before it published.