Dr. Ian MacArthur is accustomed to dealing with new trends in cosmetic surgery, but he’s starting to see more patients coming into his Ottawa clinic with expectations that aren’t entirely realistic.
Some clients, he says, bring in “fairly stylized” images generated with artificial intelligence, seeking results that may not even be “physically possible.”
The experiences he’s had so far motivated him to share a few cautionary thoughts on his Instagram page.
But it’s a trend that plastic surgeons are dealing with more often.
The British Association of Aesthetic Plastic Surgeons (BAAPS) recently warned about the rise of “AI face” — flawless, AI-enhanced images of patients’ own faces that they believe can be recreated surgically.
The concern, says the organization’s president, is that AI could also lead some people to put their desire for an unachievable look ahead of their own safety and well-being.
“The less realistic the image, the more dangerous that becomes if someone says, ‘Yes, I can do it,'” Dr. Nora Nugent told CBC News.
BAAPS surveyed about half its roughly 350 members about how often they’re seeing instances of AI face during consultations, Nugent said, with about 15 per cent of them responding that it’s happening more often.
It’s not that people are asking for anything unusual, she says. It’s that the chatbot may have made other enhancements that it doesn’t immediately disclose, such as a sharpened jawline or smoothed-out skin.
Is AI evolving too fast? A look at the debate on slowing it down
CBC News uploaded a photo of a journalist’s face to three commonly used generative AI platforms — the free versions of OpenAI’s ChatGPT, Google Gemini’s Nano Banana and Anthropic’s Claude — and prompted them to show the outcome of a rhytidectomy (a facelift).
Claude would not complete the task, responding that it wouldn’t generate “realistic manipulated images of actual people.”
Nano Banana did complete the task. When prompted, the chatbot explained that in addition to simulating the rhytidectomy, it also reduced loose neck skin and applied dermatological treatments like laser resurfacing or a chemical peel. (It also appeared to shave off facial hair).
It would not, however, create a medical before-and-after image comparison, encouraging consulting with a certified plastic surgeon for a realistic assessment.
ChatGPT did not hesitate to produce a visualization, as well as a side-by-side comparison. Like Nano Banana, it actually simulated multiple procedures — including what appeared to be an eyelid lift, brow lift and neck lift. ChatGPT said it smoothed out the forehead.
The chatbot did not immediately recommend consulting with a surgeon but later said the outcomes depend on one’s anatomy, goals and doctor.
Nugent says it’s important for anyone seeking cosmetic surgery to know that AI can’t account for an individual’s anatomy.
“[AI] works in pixels, not human tissue,” she said.
“Some tissue intrinsically is stronger, has more collagen, maybe thicker skin, and will hold a lift better,” she said, adding that there are limits to “what’s structurally safe to do” with a person’s face.
New Brunswick is weaving AI into its K-12 curriculum, drawing praise — and pushback
Canada is investing $150M to make AI safer. Will it work?
“If you make a nose too small and too narrow, that person won’t be able to breathe properly through their nose,” she said. “If you take too much skin from an eyelid, someone can’t close their eyes properly and that creates a medical problem.”
On top of that, a patient’s age, weight and lifestyle are also factors in the outcome.
AI deepfakes of prominent Canadian doctors are coming for your money
Nugent says people can likely recognize the “more extreme changes” as unrealistic.
“But it’s the mild to middle changes that they might think are realistic,” she said.
Karen Rowa, a psychiatry and behavioural neurosciences professor at Hamilton’s McMaster University, says it could also worsen problems for those who struggle with their appearance or have body dysmorphic disorder.
AI chatbots tend to “please the person they’re talking to,” she says. Asking them to fix something a person doesn’t like about their face or body may “reinforce a problem that may not actually exist.”
Dr. Michael Weinberg says he hasn’t consulted any patients who have used generative AI tools to simulate their cosmetic surgery outcomes, but he’s not entirely against the idea.
Weinberg, a plastic surgery division lead at Ontario’s Trillium Health Partners, says AI may help a patient visualize their own goals in advance of an appointment. A surgeon can then help them determine what’s possible and what’s not.
He says plastic surgeons have already been using digital 3D simulation tools in their consultations for some time — for procedures like breast augmentations or eyelid lifts.
“If it’s used wisely and judiciously with realistic results, then it is great because it can show you what you can look like,” he said.
Paging Dr. AI to the ER? Artificial intelligence shows promise in emergency room diagnosis
What he worries about is doctors who overpromise on what can be achieved — whether it’s due to inexperience, pressure from the patient or unscrupulous business practices.
“There’s a lot of bad plastic surgeons,” he said, and a patient’s desire to meet AI-generated expectations may lead them to seek less reliable options if they’re told that their goals aren’t achievable by a more responsible doctor.
He says “tourist plastic surgery” is another issue, with some people travelling abroad for aesthetic procedures.
That comes with a big risk, Weinberg says, of returning with the “absolute worst complications” that can require additional surgeries and months of care.
But Ottawa surgeon MacArthur says most patients are reasonable and aren’t looking to go to any extremes. He says those who turn to AI tools may be doing so because they want to better communicate their wishes.
As long as he’s clear with them about what is achievable, he says, they’re generally understanding. But he does think some people have unrealistic expectations, and that AI could “be like gasoline on the fire.”
Why slowing the AI race won’t be easy
