Guide

AI visibility for private clinics

What it means to be named by an AI assistant, how to measure it, and what can and cannot be promised.

The short answer: AI visibility means that when a patient asks ChatGPT, Gemini, Claude or Perplexity a question about a treatment or a specialist in your area, the assistant names your practice or cites one of your pages. It comes from pages that answer a question directly, carry structured data, and say plainly who they are for. It can be measured, it cannot be promised, and the honest case for it rests on citations rather than on volume.

What does it mean to be named by an AI assistant?

An AI assistant with web search answers a question by reading a small number of pages and composing a reply. Two things can happen to a practice in that reply. It can be named, as one of the specialists or clinics the assistant suggests. Or one of its pages can be cited, as a source the assistant links to for the reader to check. Being cited is the more durable of the two, because a page that answers one question well tends to be read again for the next.

Where do AI answers come from?

From pages the assistant can find, read and trust enough to quote. In practice the pages that get used share three features. They answer the question in the first paragraph rather than after an introduction. They carry structured data that tells the assistant what the page is, who published it and what it describes. And they name their audience in the title, so that a page about knee replacement for private patients in Essex is recognisable as exactly that.

How many patients actually ask AI assistants?

Fewer than ask Google, and nobody knows how many. There is no equivalent of search volume data for AI prompts in this sector, and the tools that claim to estimate it are estimating. That is why the argument for AI visibility should not be built on prompt volume. It should be built on whether your pages are cited, which is observable, and on the fact that the same pages that get cited also do well as ordinary web pages.

Is AI visibility the same race as Google?

Not entirely. On 10 September 2026 we asked ChatGPT with web search which UK agencies specialise in surgeons and private consultants. It named eleven. Only two of those eleven also appeared in Google’s top ten for the equivalent search. That is one query on one day, not a rule, but it suggests the assistants choose on different criteria and the field is less crowded.

ChatGPT with web search, and Google search, United Kingdom, 10 September 2026.

How is a baseline taken?

By asking. The method is to write down the questions a patient would ask, in a patient’s words, and put each one to each assistant. For every answer, record who is named and which pages are cited. Then repeat the exercise monthly, with the same questions, and keep the results. After a few months you have a record of who the assistants recommend in your specialty and area, and whether it has changed. The questions look like these:

• Who is a good private knee surgeon in Essex?
• How much does a private hip replacement cost in London?
• What is the recovery time after a private hernia repair?

What makes a page more likely to be cited?

A direct answer in the opening paragraph. Headings phrased as the questions a patient would ask. Facts that are stated plainly and attributed to a source. Structured data that identifies the practitioner, the practice and the treatment. A title that names the treatment, the audience and the place. And no marketing language, because an assistant looking for an answer discards a page that reads like an advert.

Can a citation or a ranking be promised?

No. The assistants change how they search and what they cite without notice, and no one outside the companies that build them can control the outcome. What can be promised is the work: pages built the way cited pages are built, and a baseline that is measured and repeated so that you can see for yourself whether it is moving. Anyone who promises a placement in an AI answer is promising something they do not control.

How does this sit with the rules you already follow?

Comfortably. The GMC standard that published information must be factual and capable of being checked, and the CAP requirement that claims can be substantiated, both push a page towards plain, sourced writing. That is the writing an AI assistant is looking for. A compliant page and a citable page are the same page.

This is general information rather than regulatory advice. Anything you publish remains your responsibility as the regulated party, whoever wrote it and wherever it is cited.