AI visibility for physiotherapy practices: what being named means

What counts as being named, for a physio practice?

The measurement records three grades. A mention is your name anywhere in the answer, including in a list of several or beside a caution. A recommendation is the answer putting your practice forward as suitable for the person who asked. First place is your practice being the first one the patient reads. The visibility score we report counts only the first, as the share of the twelve questions in which the practice was named; first place is reported beside it.

Each grade belongs to a question, and a patient's question to an assistant can carry a problem, a treatment and a place at once. A practice can be recommended for sports injuries and absent from every question about rehab after surgery, or named in one town and missing in the next. So the useful reading is not one number but a grid: which problems, which treatments and which places bring up your name, and which do not.

Being named also has to be right. An answer that says your practice needs a doctor's referral, when you see patients without one, turns away patients who could have booked; so does last year's fee, a location you closed or a therapist who has left. Reading every answer for the facts it states is half the work, and the facts themselves are listed in the overview for physiotherapy practices.

What a patient's question carries that a search box does not

A search box gets a few words. A question to an assistant can carry the whole situation: the problem, how long it has lasted, what the surgeon said, who it is for, what the patient can pay, when they can come. Each of those is a filter a fitting answer has to meet, and between them they sort practices by more than distance. The patient has already decided to be seen; the question is where, and that is the search intent a panel has to capture.

• The problem, in the patient's words: "lower back went out", "knee after acl surgery", "jaw clicks when i eat", "swelling after my operation".
• The treatment, when the patient already knows its name: manual therapy, lymphatic drainage, pelvic floor physiotherapy.
• Who it is for: a child, a runner, a parent after a fall, someone who cannot drive.
• The way in and the cost: "no referral", "takes my plan", "how much per session".
• The place and the hour: a town, a neighborhood, "home visits", "after work", "saturday".

None of these is unusual. Each is a fact a practice either states on a page or leaves to the phone, and a question typed on a Sunday has nobody to call.

A panel built by condition, treatment and place

The panel is the fixed set of questions put to the assistants every month, worded the way patients type. How any panel is built and kept comparable is set out in how to measure AI visibility, and the file and record format in setting up prompt monitoring. What is particular to a physio practice is the grid the questions are drawn from: the service lines you want patients for, each asked three ways.

That is twenty-one questions, more than a panel needs. Choose the rows you want patients for, take one or two cells from each, and stop at twelve, the size of a panel we run for a paying client. Once chosen, leave the wording alone: a question reworded is a new question with no history to compare against.

Two more belong above the grid. One sets your practice beside a named neighbor, "[your practice] or [other practice] for knee rehab", and shows how each of you is described. The other asks about you by name, "does [your practice] take patients without a referral", and shows what an answer states about you, fact by fact.

Word the questions as patients do. The condition comes before the treatment, the town is spelled the way locals say it, and the plan is named as it appears on the card. The same care goes by two names, physical therapy and physiotherapy, and a practice whose patients use both should ask both ways. The questions your front desk already hears, collected as in the questions physio patients ask, are the best source of that wording.

Where the panel stops: questions about the body

Leave out every question that asks for care rather than for a practice. "should i use heat or ice on my back", "is it normal for my knee to click after surgery" and "exercises for tmj" are questions about a body. Whatever an answer says to them, it is not choosing between practices, and a practice could only be named for them by publishing advice.

The test is whether a practice's name could be the answer. "physio near me for an ankle sprain that won't heal" passes, because the patient wants somewhere to go. "how long does an ankle sprain take to heal" fails, however close it sounds. When a question sits on the line, leave it out. The panel measures a choice, and the pages it leads you to fix should state what the practice does, never what a patient's body will do.

Running the panel, and reading the first result

Put each question to ChatGPT, Claude, Gemini and Perplexity, three runs each, because the same question gets different answers from one run to the next. Read each answer, record whether the practice was mentioned, recommended or first and who else was named, and keep the answer word for word. Our methodology sets out a narrower procedure for clients: the same twelve questions put to one assistant, Claude, before the fixes, once they are live and every month with monitoring, with each answer kept word for word.

Treat the first reading as a starting line. The audit behind this site's figures, 9,483 small business websites in 40 trades and 90 metros across the US, Canada, Europe and the UAE, fetched between late August and early September 2026, also put one buyer question per business to Claude with web retrieval on, for 9,240 of them. 14.9% of the businesses were named. Among the 5,928 answers that gave a clearly differentiated list rather than a hedge, the figure was 17.7%. One question, one assistant, one day, every trade together: a Claude direction, with no physiotherapy cut.

For a physio question a hedge can look like help: an answer that tells the patient to see a physiotherapist, explains what one does, and names no practice. Record it as a hedge, not a miss, because it named nobody. What to watch over the months is the share of answers that give names at all, and whether yours is among them. That first reading is your baseline measurement, and every later month is read against it.

Beside each verdict, note three things the score hides: which service line the question came from, the facts the answer stated about you, and the practices named instead. The grid shows where you are named, the facts show what to correct, and the other names show whose pages to read, as competitor analysis in AI answers does. Which of your own pages the answers cited is a further reading, and treatment pages that survive being quoted shows how to take it from the answers' sources.