We put the questions a London patient actually asks to ChatGPT, Claude, Gemini and Perplexity, and recorded which of 273 clinics each engine names. Only 7 consistently appear - the rest are invisible in the channel patients increasingly start with.
When a London patient asks AI for an aesthetic clinic, the same handful of names come back - Harley Street Skin Clinic, The Cosmetic Skin Clinic, Cadogan Clinic, PHI Clinic, Taktouk Clinic. The other 266 clinics, including most with strong reputations and busy practices, rarely surface at all. 70% of the cohort is effectively invisible in AI search.
Real, unedited answers from the study - the same handful of names recur, question after question.
Visibility = share of AI answers naming the clinic. The full ranking of all 273 clinics, with each one's band, is available on request.
| # | Aesthetic Clinic | Score | Visibility | Band |
|---|---|---|---|---|
| 1 | Harley Street Skin Clinic | 83.5 | 70% | Active |
| 2 | The Cosmetic Skin Clinic | 72.5 | 60% | Active |
| 3 | Cadogan Clinic | 63.6 | 50% | Active |
| 4 | PHI Clinic | 60.6 | 45% | Active |
| 5 | Taktouk Clinic | 55.6 | 45% | Emerging |
| 6 | House of Saab | 40.5 | 35% | Emerging |
| 7 | Dr Leah Clinics | 38.5 | 30% | Active |
| 8 | The Clinic by Dr Maryam Zamani | 35.6 | 25% | Emerging |
| 9 | Montrose London | 34.0 | 30% | Emerging |
| 10 | 111 Harley St. | 33.9 | 25% | Active |
| + 263 more clinics ranked, with each one's visibility and band. Get the full leaderboard & where you rank → | ||||
The top 5 are named in 45-70% of answers. By rank 10, visibility is already down to 25% - and 263 clinics sit below this table.
These directories, regulators and platforms surfaced in AI answers. They are not competing for the patient - they are the third-party sources AI re-reads. Getting listed and well-reviewed on them moves your own visibility.
| Source | Type | Engines |
|---|---|---|
| marieclaire.co.uk | Publication | 3 |
| treatwell.co.uk | Directory | 2 |
| booksy.com | Directory | 1 |
| dermatologists.london | Directory | 1 |
| injectablesbooking.co.uk | Directory | 1 |
| + more sources cited across the cohort. Get the full source list → | ||
When ChatGPT, Claude, Gemini or Perplexity answer a patient's question, the response is assembled from six technical inputs. No single marketing discipline covers them all - which is why 70% of clinics rarely surface.
We cover all six layers - the audit, the fixes, and the ongoing visibility - so AI starts recommending you. One programme, every engine.
28 Labs measures how brands get recommended across ChatGPT, Claude, Gemini and Perplexity. We've analysed 11,000+ AI answers covering 100+ brands across 15 markets - a uniquely deep view of what AI actually recommends, and why.
Harley Street Skin Clinic, The Cosmetic Skin Clinic, Cadogan Clinic, PHI Clinic, Taktouk Clinic are the most consistently named across ChatGPT, Claude, Gemini and Perplexity. Only 7 of 273 clinics appear regularly.
Just 7 of 273 clinics consistently appear; 70% rarely surface at all.
The most-cited sources are marieclaire.co.uk, treatwell.co.uk, booksy.com - the directories, regulators and platforms AI re-reads to build its answers.
Asked across the four engines, AI most often names Harley Street Skin Clinic, The Cosmetic Skin Clinic, Cadogan Clinic. The full ranking, and where each clinic sits, is on this page.
Each clinic is scored on whether ChatGPT, Claude, Gemini and Perplexity name it across the questions patients ask, May 2026. Citations reflect the retrieval / web layer; some recommendations draw on training data with no visible citation.
This page shows the public top 10. Get the full ranking of all 273 London clinics, exactly where you sit across all four engines, and the specific moves to climb.
Press & data requests: [email protected]
How we did it. Patient questions x 4 engines (ChatGPT, Claude, Gemini, Perplexity), 273 London clinics scored on whether each engine names them. "Visibility" = share of answers naming the clinic. Citations reflect the retrieval / web layer; some recommendations draw on training data with no visible citation. Data captured May 2026.