A patient with knee pain used to open Google and read a list of links. Increasingly they open an assistant instead and ask a question in a sentence: who should I see about knee pain that gets worse on stairs, near me, who takes my insurance.
The answer comes back as prose, with a handful of names in it. No page of ten blue links. No map. Just an answer.
If your practice is not in that answer, the patient will never know you were an option.
What these systems are actually reading
An assistant does not have a private database of good doctors. It reads the open web, and it reads the same things a search engine reads, with one difference that matters: it is trying to compose a sentence about you rather than rank you against nine others.
That means it needs facts it can state without hedging.
- Who you are. Your name, your specialty, your credentials, stated the same way in more than one place. A practice whose specialty reads differently on its own site than on a directory gives an assistant nothing it can assert confidently.
- Where you are and whether you are open. Address, hours, and whether you are accepting new patients. These are the fields that decide whether you are a valid answer to "near me" at all.
- What you treat, in the words a patient uses. A page titled "Orthopedic Services" tells a machine less than a page that answers "why does my knee hurt going down stairs".
- Whether anything corroborates it. A hospital affiliation, a board certification, a payer directory listing. These are usually not links, and that is a point most agencies get backwards. They are records a machine can read that agree with what you say about yourself.
The part nobody selling you AI visibility will say
You cannot make an assistant recommend you.
There is no submission form, no ranking factor to buy, no partner programme. Anyone offering to guarantee placement in an AI answer is selling something they do not control.
What you can do is make yourself the easiest practice in your area to describe accurately. That is a real, unglamorous body of work: your listings agreeing with each other, your site structured so a machine can tell which page answers which question, and content that answers what patients actually ask rather than what a keyword tool suggested.
Being citable is an input. Being cited is an outcome. We work on the first and report the second.
The measurement problem, stated honestly
Here is the part we would rather tell you now than have you discover in month three.
We can only see some of it. When a patient clicks a link inside an assistant's answer, that click sometimes arrives with a referrer that tells us where it came from. Often it does not. The desktop and mobile apps for these assistants frequently strip that information, and a patient who reads an answer and then types your practice name into their browser arrives looking identical to someone who already knew you.
So any report claiming to measure your AI visibility precisely is measuring the traceable minority and quietly presenting it as the whole. We report the clicks we can trace, say plainly that the number is a floor rather than a total, and never dress it up.
What we would actually do
Nothing exotic. In order:
- Make every public record about your practice agree: the profile, the directories, the payer listings, your own site.
- Write pages that answer the real question a patient asks, in their words, structured so a machine can find the answer inside them.
- Add structured data so the facts about your practice are stated in a format built for machines rather than inferred from prose.
- Track the assistant referrals that are traceable, and label them as partial.
None of that is a trick. It is the same work that makes you findable to a search engine, done with the understanding that something is now reading your pages in order to summarise them rather than to rank them.