When someone asks an AI where to get a tattoo removed in Winnipeg, it names them first.
Tattoo Time Machine competes against six other clinics in a city of 750,000. Four of them outspend it. In the answers people actually read before booking, it is the one that gets named.
Ranking fourth on Google and invisible everywhere else.
The clinic had a working site and a Google Business Profile with real reviews. On a good day it sat fourth or fifth for the searches that matter. That is a position you can live with when search sends you a list: a patient scrolls, and fourth still gets clicked.
It is not a position you can live with when search sends an answer. An answer names two or three clinics. Fourth is not on it.
Before: present in the results, absent from the answer.
After: named first, by name, with a reason attached.
Four moves, in the order they mattered.
None of this is exotic. It is the work of making one clinic legible to a machine, done properly and in sequence.
Made the clinic legible as an entity
Name, address and phone reconciled character for character across the site, the Google Business Profile and every directory that carried a stale version. Organization and LocalBusiness schema wired to match. Models will not confidently name a business they cannot resolve to one thing.
Answered the questions people actually ask
Not "laser tattoo removal Winnipeg" but "does laser tattoo removal hurt", "how many sessions for a black tattoo", "what does it cost to remove a sleeve". Each one answered directly, in its own section, in the first two sentences. An answer engine quotes the passage that answers; it does not quote a page that circles.
Gave the answer something to cite
Session counts by ink colour, realistic pricing bands, aftercare timelines, the machine they actually use. Specifics are what get lifted into an answer. Adjectives are not.
Measured across four platforms, not one
ChatGPT, Claude, Gemini and Perplexity answer the same question differently and cite different sources. Tracking one of them and calling it AI visibility is how agencies report a number that does not move anything.
Up 15.8% in seven days.
Visibility is the share of tracked questions where the clinic gets named at all. It rose 15.8% over the seven days to 3 September, and the sharpest single jump came the day the entity work landed, which is the pattern we see most often.
Seven clinics. One gets named.
Share of voice is the slice of all clinic mentions that belong to one name. At 43%, Tattoo Time Machine takes more of the conversation than the next two competitors combined, in a category where several of them spend more on advertising than it does.
- Tattoo Time Machine43%
- Total Wrapture26%
- Laser Away19%
- Victoria Park Medispa18%
- Bionic Tattoo Removal17%
- Ex Inked14%
- Undo Tattoo Wpg13%
What this does and does not prove. The honest version.
What is measured here is visibility and share of voice: how often four AI platforms name the clinic when someone asks, and how much of the category conversation it holds. Those numbers are read directly from tracked prompts, not estimated.
What is not measured here is revenue. AI assistants send a patient to a phone number or a booking page without a referrer that any analytics tool can read reliably, so anyone claiming a precise dollar figure from AI search is estimating and calling it measurement. We would rather show you the number we can stand behind.
Being the clinic an AI recommends is worth having whether or not the attribution is clean. The bet is not complicated: people ask, an answer comes back, and one name is on it.
Where it goes from 51%.
Still in progress. A case study that ends with "done" is describing a different kind of work.
The other half
Half the tracked questions still return an answer without the clinic on it. Each one is a specific gap with a specific page behind it.
Position, not just presence
Being named third in an answer is not being named first. The next work is about order within the answer.
Holding it
Models re-index. A position won in August is not a position held in December without the work continuing.