MediSearch · Answer engine optimisation for healthcare
When someone asks an assistant who to see for their knee, they get an answer above the links. Either your clinic is named in it or it is not, and many people stop reading there. MediSearch measures which one is happening, market by market, and ranks what to work on.
Search engine optimisation asks: where do we rank? It assumes a page of results and a patient willing to work through it. That assumption held for twenty years and it is quietly expiring.
Answer engine optimisation asks something narrower and harder: when an assistant answers this question, are we in the answer, and what does it say about us? The answer is one paragraph that names a few clinics. In the 71,000 answers we have recorded, the typical one names four.
Ten links. The patient chooses.
You compete for position. Being fourth still puts you on the screen, and a patient who scrolls can still find you.
Measured in rankings, impressions and clicks, all of which assume a link was shown and could be clicked.
One answer. The assistant chooses.
You are named or you are absent. The answer sits above the links, and many people stop there, so being named in it is a different result from ranking under it.
Measured in whether you were named, whether you were recommended, and what was said about you when you were.
“Best physiotherapist” has no single answer. Asked from one location it names one set of clinics; asked from the next it names another. A clinic with six locations does not have a visibility problem: it has six of them, and they are usually not the same problem.
This is why a single national score is close to useless for a practice with more than one site, and why MediSearch measures every subject against every market you serve rather than reporting one number.
Most visibility work is about being found. In healthcare there is a second failure that is worse: being found and described incorrectly. An assistant telling a patient that you require payment upfront, do not take their insurance, or do not treat their condition is a claim about your practice, delivered with confidence, to someone deciding where to go.
An assistant steering someone to a competitor is not a lost impression. It is a person who needed care and went somewhere else, and neither of you will ever know it happened.
We write the questions your patients ask (about conditions, about treatments, about you by name) and ask them in Google's AI Overviews, from each market you serve, on a schedule. Then we read every answer.
Illustrative · not a real clinic
| Overall | Riverside | Eastport | Lakefield | Northgate | |
|---|---|---|---|---|---|
| Health concerns | 46% | 72% | 58% | 44% | 9% |
| Treatments & services | 48% | 70% | 38% | 66% | 18% |
| Your name | 77% | 91% | 78% | 84% | 55% |
Read a column and you have one clinic's standing. Read a row and you have one kind of question across the group. Northgate is not a marketing problem in general: it is invisible on the conditions patients actually search, while people who already know the name still find it.
That grid is the difference between “our AI visibility is 57” and knowing that Northgate needs condition pages and Eastport does not.
Written from your services, your conditions, your markets and how your patients pay rather than a generic keyword list. The same questions are reused every scan, so the numbers stay comparable over time.
Scans run on a schedule, in Google's AI Overviews. One reading is a snapshot; the value is in the line it draws once you start changing things.
Every finding is scored by an estimate of how much it would move the numbers and how easily it can be done, and grouped so three versions of the same job read as one. It opens onto the exact questions it came from.
Reports built entirely from real scan data. No model writes them, so a report cannot claim something the dashboard does not show.
They overlap and they are not the same. Good SEO helps, because assistants read the open web, and a page that ranks well is a page they are more likely to retrieve. But ranking first for a term an assistant never surfaces is worth nothing, and being named in an answer without ranking at all is worth a great deal. The work is also different: much of what AEO fixes is factual record rather than content volume.
It is early, and worth saying so plainly. It is established enough that Gartner published a 2026 Market Guide for Answer Engine Visibility Tools, and that the large SEO platforms now ship AI visibility modules. MediSearch is built around geography, asking every question from every market you serve, which for a multi-site clinic is the whole question.
Every rate comes from answers we recorded and can show you. Click a percentage in the grid and you get the questions behind it and how each one did. The one blended number, the score, says how it is built, and anything estimated is labelled as an estimate.
The first scan gives you the grid above for your own practice, filled in with real answers. That is usually enough to know whether this is a problem worth your attention.
MediSearch measures how AI answers the questions your patients ask. Built for multi-location healthcare practices. Back to the front page.