For med spas
Some San Francisco med spas were named by no engine at all.
On 29 July 2026 we put a set of buyer questions to ChatGPT, Gemini and Perplexity with the same pool of San Francisco med spas. Most came back at least once. Behind those names sits a different booking system almost every time.
- Written for
- Med spas and aesthetic clinics
- Measured
- 29 July 2026, San Francisco
- Booking systems in the pool
- A different one behind almost every door
- Connector confirmed
- None in the pool
What an agent sees at a connected med spa
- 11:00 AM$450
Botox with Dr. Lee
30 units. Quoted per unit after an assessment.
- 2:00 PM$180
Hydrafacial with Nina
60 minutes. Open, and held by nobody until a booking is confirmed.
- 4:30 PM$900
Filler consult
45 minutes, with the treatment booked separately.
- 5:15 PM$220
Laser touch-up
20 minutes, and the last slot an agent can see today.
We read the med spas of San Francisco one by one and found a different booking system behind almost every door, plus sites where we could not tell what runs the calendar at all. Square Appointments, the one system we support, was not among them. For a med spa the honest offer on this page is the queue.
What the answer engines said about this market in one day
Every query went out once per engine with the whole pool named as the field, so a single call scored every business at once. No call failed.
- 01A minority of the pool was named by every engine. Most of what came back came from a single engine.
- 02Even the widest reach in the pool left most query and engine slots empty. Where one practice runs several addresses on one domain, those addresses share the same reading.
- 03ChatGPT and Perplexity named the same share of the pool. Gemini named fewer.
- 04The queries were the ones a buyer types: same day botox appointment san francisco, book a facial in san francisco this week, chemical peel san francisco appointment, medspa marina district san francisco.
- 05The practices with the widest reach ran a booking page that resolved to no known platform, and the practices sharing one platform were barely named at all.
What this run does not show
The limits are recorded in the run file itself and travel with every finding taken from it. Anyone quoting this page carries the date and the method along with it.
- 01One API run on one date. AI answers are non-deterministic and personalised, and a rerun returns a different reading.
- 02These calls went through provider APIs: gpt-4o-mini with the OpenAI web search tool, gemini-3.5-flash with Google Search grounding, and Perplexity Sonar. The consumer apps run other models, other prompts and personalisation.
- 03No location signal was sent. A real San Francisco user is geolocated; these calls were not.
- 04A mention is a domain or brand-name match in the answer text or in a citation, so it can over-count on generic names and under-count when a business is named without its domain.
- 05Positions come from citation order. No engine publishes a ranking here.
The answer names you, then the calendar has to answer
A citation puts a name in front of somebody who has already decided to book. The step after that runs against the booking system, and for this market that step is where things stop.
- 01An agent reading a marketing page can offer a phone number and a link, and the buyer finishes the job by hand.
- 02Holding a real time needs a system that answers a machine with services, staff, length and open slots.
- 03Payment is settled the way it is settled today, and the agent path does not complete it.
- 04When the pool was assembled, no business in it had a confirmed connector: most had a provider we could identify, and the rest published too little to tell.
A different booking system behind almost every door
The pool was collected on 22 July 2026 by reading each booking link and the provider script. No system runs this market: the systems came back almost as varied as the pool itself, and a share of the sites resolved to no known platform.
- 01Even the most common system in the pool carried only a handful of practices, and none carried more.
- 02Most systems carried a single practice each, and some sites published too little for us to tell what runs the calendar at all.
- 03Square Appointments, the one backend where availability reads are live today, appeared nowhere in this pool.
- 04For a med spa on any of these systems the answer today is that there is no path yet. Naming yours is what puts you on the list we write to.
What a med spa sells, and what an agent can price
Buyers type units, areas and named treatments, and the queries in this run show exactly that. A practice answers with a consultation and a licensed provider.
- 01Injectables are quoted by unit or by area after an assessment, so a fixed price on an agent-visible menu commits the room to a number nobody has checked.
- 02Laser and energy treatments depend on skin type, history and course length, and the first appointment is usually the assessment itself.
- 03The provider matters: a nurse, a nurse practitioner and a physician are separate bookings on the same calendar, each with its own intake.
- 04The version that works is publishing the consultation as the bookable service and keeping the treatment off the agent-visible menu.
- 05Where a facial or a peel already runs to a fixed length and a fixed price, that is the first thing worth exposing.
Where your own system stands
None of the systems in this pool has a path today. Naming the system you run is what puts you on the list we write to the day that changes.
To score your own setup, start with the AI readiness check, and the argument without the trade language is on the page for service businesses.
Supported system review
Tell us what runs your calendar and you get one of three answers: agents can read your availability today, your system is in verification, or there is no path yet and we write to you the day that changes.
Questions we get
Our practice was named by nobody. What does that mean?+
For those queries, through those APIs, with no location signal, none of the engines surfaced you on the day of the run. It is one reading of one day. The work it points at sits upstream of us: the pages, the listings and the structured detail an engine can quote.
Can WhatsDo connect the system our practice runs?+
Only if it is Square Appointments. That is the supported path: availability reads are live there and the booking write runs in closed beta against a controlled merchant. Every other system we found in this pool is on request, we publish that, and we name no date for changing it.
Would being named more often bring us bookings?+
We publish no forecast and no practice should accept one from us. The run measures whether a name appears in an answer. What that turns into depends on the step after the answer, and for every business in this pool that step is still a phone call, a form or a walk-in.
Can an agent book botox or a laser course?+
It should not, and the setup we recommend keeps it from trying. The bookable object is the consultation. A course, a unit count and a price come from a provider who has seen the patient.
Why measure ChatGPT, Gemini and Perplexity through APIs?+
Because it is repeatable and it names its own model. Every finding here carries the engine, the model, the search tool and the date, so a rerun in a month is comparable to this one. What a person sees in the consumer app on their own phone is a separate measurement, and this run stands only for itself.
Find out today where your practice stands with the engines.
Nothing changes for your staff, and you can disconnect at any time.
Building an agent? Start here