Almost everything currently sold as AI in aesthetics is a camera and a marketing budget.
The consultation simulators are the clearest example. A patient uploads a photograph, a model generates a version of their face with a smaller nose or a fuller cheek, and the clinic calls it planning. It isn't planning. It's a rendering of an outcome that nobody has committed to delivering, produced by a system that has no knowledge of the patient's anatomy, skin quality, healing history, or the actual mechanical limits of the procedure. It sets an expectation and transfers none of the risk of meeting it.
I'd go further: these tools are actively harmful in consent. A patient who has seen a photorealistic image of their post-operative face has been shown something. What they were shown cannot be guaranteed, and once seen it is very difficult to unsee. The gap between that image and the real result becomes the surgeon's problem at six weeks.
Where it genuinely earns its keep
There are places where the technology works, and they're less glamorous.
Documentation is one. Standardised photography with automated pose and lighting correction solves a real problem, which is that most clinical photography in this field is inconsistent enough to be useless for comparison. If you cannot reliably compare a patient to themselves a year ago, you cannot audit your own outcomes, and most practitioners can't.
Triage is another. Systems that flag whether a presenting complaint is likely to be within the scope of a non-surgical intervention are useful — not to make the decision, but to make sure the conversation starts in the right place.
And there's an emerging use in complications: pattern recognition across large numbers of adverse events, which no individual practitioner sees enough of to learn from. That's the application I'd most like to see funded, and the one with almost no commercial pull behind it, because nobody sells a subscription on the strength of the problems it prevents.
The pattern
The tools that help are boring and operational. The tools that sell are visual and speculative. A field that has spent a decade learning to distrust its own marketing should be able to tell the difference by now.