AI Edited My Face Without Asking. At First I Laughed It Off—Then I Started Seeing Myself Differently

“Claude, fix my face.” That was not the prompt that I, or rather, my boyfriend, gave an AI platform, but it decided to enhance me anyway. A few weeks ago, my partner was taking a photo of me when a woman strolled past, walking a dalmatian. “I should've asked to borrow her dog,” I quipped, as I too was bedecked in spots. “Stick out your arm like you're holding a leash,” he replied. I didn't give it much thought, but within minutes, he'd uploaded the photo to an AI platform, and back came a rendered image of me holding a dalmatian.

He presented the image to me, perhaps expecting me to be impressed by the speed with which this relatively new tech could update my image when, only a short while ago, it would have taken someone experienced in Photoshop to create it. But the first thing I noticed wasn't the dog; it was my face.

Without prompting, the AI had given me a beaming smile with perfectly straight, white teeth. My cheekbones were higher, my face smoothed out, my eyebrows better groomed. It still looked like me, but I guess the intention was to show a 'better' version of me, when all that had been requested was for a spotted dog to be added in.

Later, when I told friends about this odd experience, they suggested I prompt the AI further, turning me into Cruella de Vil. What came back was an even more yassified version of me: the smile tighter, the skin tauter, even if a dodgy wig did crown the look.

I thought again of this incident, when Meta's latest AI rollout, which has since been pulled, prompted a flurry of concern. During my afternoon doomscroll, my feed became awash with the news that users would be able to generate AI images using other people's likenesses, with news reports rushing to show people how to opt out. The fury of users was so widespread that Meta made the unusual decision to reverse this feature; posting on their newsroom that: "We've heard the feedback that this feature missed the mark, so it’s no longer available." Much of the conversation rightly centred on consent, ownership and the implications of AI-generated versions of ourselves.

But my mind kept returning to a different question, because before AI can manipulate our images, it first has to decide what we should look like. I thought back to yet another conversation I'd had, this time with Nikki Lilly, who told me what it feels like to have strangers on the internet use AI to manipulate your face. Again, supposedly to "fix" her, or, more specifically, her facial difference.

Though the severity and mental health implications varied wildly, in both cases, no one had asked for a makeover.

For my part, there was no prompt requesting whiter teeth, smoother skin or more sculpted cheekbones. Yet the AI seemed to assume those were improvements worth making anyway. It wasn't simply following instructions; it was making aesthetic decisions on my behalf.

That might sound trivial. After all, we've spent years voluntarily downloading beauty filters and smoothing apps. But there's an important difference between choosing to alter your appearance and having an AI decide to do it for you, especially given that there's an entire cottage industry sprouting up which allows users to upload their image to AI platforms that will then suggest—and later sell—beauty tweakments. Qoves, a facial aesthetics platform that uses AI and computer vision to analyse users' facial features and provide personalised appearance assessments and recommendations based on aesthetic research, has 1.1 million followers on Instagram alone.

The more I spoke to experts, as well as friends who’d had similar experiences (unkempt eyebrows seems to be an Achilles heel for AI), the more I realised my experience wasn't just an odd quirk of one platform. It points to a much bigger question about the beauty standards being baked into the AI tools we're increasingly relying on to create, edit and represent us. If these systems are learning from the internet's existing ideals of attractiveness, are they also reinforcing them—one automatically enhanced image at a time?

​According to psychologists, that's exactly what makes these seemingly minor edits so powerful. "When people repeatedly see a slightly enhanced version of themselves—smoother skin, more symmetry, subtly narrowed features—it can recalibrate their internal standard of what they 'should' look like," explains Tara Well, associate professor of psychology at Barnard College of Columbia University. "Over time, the unedited face can begin to feel like a deviation rather than the baseline."

We've become used to talking about social media through the lens of comparison. For years, experts have warned that scrolling through impossibly beautiful celebrities and influencers can leave us feeling inadequate. But AI introduces a new, more intimate–and insidious–comparison: instead of measuring ourselves against other people, we're measuring ourselves against idealised versions of ourselves.

Where traditional beauty filters require an active choice–we know we've selected one, and we're usually aware that what we're looking at isn't entirely real–AI-generated enhancements are different. They're often subtle, automatic and, in my case, completely unsolicited.

"Traditional filters involve a degree of choice and awareness," says Well. "Automatic enhancements blur that boundary. Because the changes are subtle and default, users may not fully register that their appearance has been altered, which can make the enhanced version feel more 'real'. The less effortful and more invisible the modification, the more psychologically persuasive it becomes."

"In short," says Well, "these systems don't just reflect beauty standards—they quietly standardise them. And because the changes are subtle and automatic, they may be more powerful than traditional filters in shaping how people see themselves."

If AI is discreetly reinforcing conventional beauty ideals, the next question is where those ideals came from in the first place.

Professor Clare McGlynn, an expert in law and gender-based violence, argues that these technologies don't exist in a vacuum. They are trained on the same cultural biases that have long shaped women's relationship with their appearance.

"Your example shows just how deeply embedded traditional views of women and beauty are," she tells me. "We know social media exposure has adversely impacted girls' self-esteem and body image, contributing to psychological distress and self-harm. The concern is that AI could intensify those harms because it makes these 'improvements' feel even more seamless and inevitable."

In many ways, that's what unsettled me most. It wasn't that the AI had made dramatic changes to my face. Quite the opposite. The edits were subtle enough to feel plausible. The kind of tweaks you might struggle to put your finger on (worryingly, my partner hadn’t even noticed), but that, combined, nudged me a little closer to a familiar beauty ideal.

I’m not the only person left questioning myself. Journalist Kish Lal had a strikingly similar experience after asking Claude to analyse her appearance as part of a reporting project. Like me, she thought she'd be able to keep an emotional distance from the results. Instead, she found herself dwelling on critiques she'd never previously considered.

"I'm an AI hater to my core," she laughs. "I hate the way it smoothes and flattens whatever it touches." Even so, one suggestion lingered: again, it was the eyebrows. "The robot told me my eyebrows would look better if they were more manicured," she says. "I'd always loved my bushy brows. I used to rip photos of Natalia Vodianova out of magazines because of hers. I even joked to a friend that the AI must just be telling everyone to thread their eyebrows."

Then her friend tried the same experiment. "It complimented her brows and told her to get filler instead," Lal recalls. “I couldn't believe that my theory was not only debunked, but that I perhaps, maybe I could look better if I got my eyebrows threaded regularly. I scheduled an appointment to get them done the next day.”

It's easy to dismiss moments like these as trivial or even funny. But they reveal something much more uncomfortable. AI doesn't need to tell us we're unattractive to influence how we see ourselves. Sometimes, all it has to do is quietly suggest that we'd look a little better if we were just... slightly different.

The more sophisticated these tools become, the more they're marketed as neutral assistants—helping us create images, edit photos or generate avatars in seconds. But assistants aren't neutral if they're making value judgements on our behalf.

On their own, those edits seem almost laughably insignificant. I joked about my 'glow-up' with friends the way Kish Lal did and yet in both instances, we were left wondering what difference those subtle suggestions would make to our appearance.

As AI-generated images become more common—and companies race to build tools that can create ever more convincing versions of us—we'll spend a lot of time talking about consent, copyright and misinformation. We absolutely should, but I'd argue we also need to ask another question: when AI decides what we should look like, whose idea of beauty is it reproducing?

Because if I ask it to add a dalmatian, I'd quite like it to stop there.