Patients now ask an AI before they search. Here is what the 2026 data says, and what it changes for a private clinic.

The line was crossed in the first half of 2026

rater8's third Patient Choice Report, published on 3 June 2026 from a survey of roughly 1,000 US adults, found that among people who had searched for a doctor in the previous year, 36 percent cited AI tools as an influence on their choice. Google search was cited by 34 percent and doctor referrals by 32. A year earlier the AI figure was 17 percent. Two details matter more than the headline. Adoption was highest among people aged 45 to 60 (64 percent) and 30 to 44 (52 percent), which is the facelift, implant and hair-transplant demographic rather than the teenage one. And 60 percent said they trusted the AI summary without checking it, even though 66 percent reported having met wrong information about a provider in one.

BrightLocal's Local Consumer Review Survey, published in March 2026, measured the same shift from a different angle: 45 percent of US consumers said they use AI tools for local business recommendations, up from 6 percent a year earlier. Ninety-seven percent of them still cross-check against reviews, and 64 percent said they trust the AI recommendation as much as they trust reviews.

The UK has no equivalent patient survey of the same quality yet, which we note rather than paper over. The mechanism is identical, the tools are the same, and OpenAI reports that health questions run at around 40 million a day on ChatGPT globally, 70 percent of them outside clinic hours. We treat the US figures as the best available proxy for London and say so when we present them.

What the answer is made of

Whitespark's study of AI Overviews in local search, reported by Search Engine Journal in July 2026, found an AI Overview on 68 percent of local searches and on 92 to 97 percent of informational and hybrid ones, against a local pack on 39 percent. More important for a clinic is where the answer comes from: 77 percent of what the AI cited for a local or branded query was off-page. Reviews, Reddit threads, YouTube explainers and "best clinics in [city]" listicles carried the answer. The business's own website accounted for roughly a quarter.

Engines also disagree with one another. Citation overlap between ChatGPT and Perplexity has been measured at about 11 percent (single-source), and ChatGPT is reported to swap roughly three quarters of its cited sources week to week (SISTRIX). Each engine is a separate front, and a single screenshot proves nothing. This is why we measure with a prompt panel, repeated, and report a trend rather than a position.

Why it matters more for clinics than for most businesses

Three features of private clinic economics make the shift sharper than average. The purchase is high-value and high-anxiety: a facelift at $12,000 to $19,000, a full-arch implant case at around $15,000 per arch, a hair transplant at $8,000 to $15,000 in the US and UK. Patients rehearse these decisions at length, and increasingly in conversation with an assistant. The category is cash-pay, so discovery is not mediated by an insurer or a referral pathway. And the answer names three to five providers where a search page listed ten or more. The practice is in the answer or it is not.

Meanwhile the cost of the old route keeps rising. Healthcare search advertising benchmarks for October 2024 to September 2025 put general dentistry at $7.03 per click and $84.77 per lead, orthodontics at $8.76 per click, and plastic surgery at $102.51 per lead (LocaliQ). Blended cost per acquired patient runs to about $610 in plastic surgery, the highest in medicine (Dentplicity, ThinkBasis, 2025 to 2026).

What to change, in order of payback

1. Know whether you are named

Ask twenty questions a patient would ask, across three engines, three times each, and record who is named. Most clinics have never done this and most agencies do not report it. It takes an afternoon and it is the first point of our free Business Scan.

2. Fix what the engines read

Because three quarters of the evidence is off-page, the work is mostly unglamorous: a steady, consented review programme with every review answered; clinician credential pages consistent with the registers and society profiles that list them; a complete Google Business Profile; presence in the directories, forums and video that engines cite for your procedures. Around four fifths of this is search and reputation fundamentals done properly. One fifth is new: content shaped to be quoted accurately, entity consistency, and measurement.

3. Answer the question patients are afraid to ask

Cost is the most common unanswered question on clinic websites and one of the most common prompts patients put to an assistant. A procedure page that explains what happens, who performs it, what the risks are, how pricing works and what a consultation involves is both what a careful patient wants and what an engine can cite. It is also permitted as information in every market we work in, where advertising the same procedure often is not.

What the data does not yet show

Absolute AI-referred traffic to clinic websites is still small; the case rests on the growth rate, on the winner-take-most shape of the answer and on the quality of the visitor (one vendor study puts AI-referred visitors at 4.4 times the conversion rate of organic, single-source). Health answers are also conservative: engines favour institutional and clinical sources for anything that looks like medical advice, which rewards clinician-authored information and punishes marketing copy. A clinic that publishes like a clinician, not like an advertiser, is on the right side of that.