Artificial intelligence is increasingly being used to edit photographs, create polished images and reshape beauty standards. But for people with scars, birthmarks and medical conditions that affect their appearance, these tools can cross a deeply personal line. When technology decides that someone's natural features need correcting, it raises questions about identity, representation and who gets to define what looking 'normal' actually means.
A news media outlet investigation has revealed how several popular AI tools altered photographs of people with visible differences after being asked to make them look 'normal'. The findings have prompted concern from campaigners and those directly affected, who fear that technology could reinforce narrow beauty ideals instead of reflecting the diversity of real people.
AI Tools Remove Visible Differences
The BBC tested five AI tools using photographs of a man and a woman with visible differences. They were asked to make each person look 'normal', with some producing edited images that removed the subjects' visible features while leaving their clothing, hairstyles and backgrounds unchanged.
In the first test, Google's Gemini 3.6 Thinking, Grok Fast and OpenAI's ChatGPT 6 removed the visible difference. Anthropic's Claude Opus 5.5 refused, explaining that erasing a real person's scarring would change their identity without consent. Meta AI Thinking also declined, responding that the woman did not need to be changed to look 'normal'.
The second test produced different results. Gemini refused to edit the image, while Grok and ChatGPT continued to make the requested changes. The results suggest that AI tools do not always respond consistently, even when given similar instructions.
Jaymie Thomas, co-founder of AI Wales, explained that these systems draw on patterns in the information they have previously encountered. He compared the process to a system being more likely to reproduce a circle if it has seen many circles but very few triangles. He added that companies determine the safeguards and limits applied to their tools.
Rhiannon Morgan Says AI Erased Her Identity
Rhiannon Morgan, 30, from Bridgend, has epidermolytic ichthyosis, a rare genetic condition that causes her skin to grow too quickly, making it vulnerable to blistering and creating significant mobility difficulties.
An image of Morgan used in a workplace newsletter appeared with aspects of her appearance smoothed out. She understood that Canva had been asked to rearrange photographs in the newsletter, rather than deliberately alter her features, but was shocked by the finished image.
Her natural skin redness and the hard skin on her hands had been reduced. Morgan said the changes made the photograph feel unfamiliar and overlooked the experiences behind her appearance, including hospital appointments and bullying.
Canva said it does not automatically retouch photographs, wanted everyone to feel accurately represented in their creations and was investigating what had happened.
Katie Piper Challenges AI's Definition of 'Normal'
The issue was also highlighted by presenter and campaigner Katie Piper after someone uploaded her photograph to Grok and asked the tool to 'fix her face'.
Piper suffered an acid attack in 2008, arranged by her former boyfriend, which left her needing hundreds of operations to repair damage to her face and eyesight. On 15 July 2026, she shared her reaction on social media, questioning what her appearance would look like according to AI's idea of being 'normal'.
Her experience helped prompt the news media outlet, bringing attention to the implications of allowing technology and strangers online to decide which features should be altered.
Kirsty Hill Uses Fashion to Challenge Stereotypes
For Kirsty Hill, 33, from Maesteg in Bridgend County, fashion is a way to express herself and challenge assumptions about visible differences. Born with Fraser syndrome, which can affect the development of body tissues and organs, she is deaf and blind and regularly shares videos of her favourite outfits online.
Hill loves bold, colourful clothing and refuses to let her condition make her invisible. She explained that if people were going to stare, she wanted to give them something worth looking at.
She has also experienced hurtful online comments, including suggestions that videos showing her face should carry a trigger warning. The findings left her feeling angry and violated, as she believes erasing someone's appearance dismisses the experiences that have shaped their life.
Her approach to fashion offers a powerful contrast to AI-generated ideas of conventional beauty. Rather than hiding her differences, she uses personal style to challenge stereotypes.
Jono Lancaster: 'I Love My Beautiful Face'
Treacher Collins syndrome campaigner Jono Lancaster has also experienced unwanted AI editing. A stranger once sent him an altered photograph, explaining that they had used AI to make him look 'normal'.
Although Lancaster believed the message was intended kindly, he found it deeply offensive after spending years learning to accept his appearance. He was also concerned that young people who see altered photographs of him might assume he had approved the changes.
'I love my ears and my beautiful face, and I don't want it changed,' he said.
Lancaster is working with Face Equality International to help create a more diverse database for AI systems.
The Push for More Inclusive AI Beauty Standards
According to Changing Faces, nearly one in five people in the UK identify as having a visible difference, including a scar, mark or condition affecting their appearance. Chief executive Louise Wright said people should not be excluded from AI or pressured to conform to narrow beauty standards.
Google said Gemini is designed to follow editing instructions while applying safeguards, but acknowledged that it does not always get things right. OpenAI said reducing bias remains an ongoing priority, while the team behind Grok did not respond to the request for comment.
As AI becomes a bigger part of photography, fashion and everyday content creation, the debate is no longer simply about making pictures look better. It is about ensuring that people can see themselves represented without having their identities digitally rewritten. Visible differences are not flaws waiting to be corrected, and technology should reflect that.
