If you scroll through Instagram or TikTok for a few minutes, you might notice something strange. A lot of people online look almost too perfect. They have smooth skin, perfectly shaped noses, big eyes, and symmetrical faces. But sometimes those faces are not even real. Many of them are created using artificial intelligence.
This raises an important question. If AI is creating the images we see online, who is actually deciding what “pretty” looks like?
AI image generators learn by analyzing millions of pictures from the internet. These pictures come from social media, advertisements, movies, and stock photos. The AI studies patterns in these images and learns what kinds of faces appear the most. Then it uses those patterns when it creates new faces.
The problem is that the internet already has strong beauty standards. If most of the pictures online show young people with clear skin and certain facial features, the AI will learn that those features are what people consider attractive. As a result, the images it generates often look very similar.
Some researchers have found that AI-generated faces often show lighter skin tones and younger-looking features more frequently than others. This shows that AI is not neutral. It reflects the data it was trained on. If the data has bias, the AI will likely repeat that bias.
Another issue researchers discuss is called “algorithmic lookism.” This means that AI systems sometimes connect attractive faces with positive qualities such as intelligence, kindness, or success. If technology keeps reinforcing these ideas, it can strengthen the belief that appearance determines someone’s value.
Social media filters also play a role. Many apps now use AI filters that smooth skin, change face shape, or adjust facial features to make people look more “ideal.” Even though the changes can seem small, constantly seeing these edited images can slowly change what people think normal faces should look like. For teenagers especially, this can affect confidence and self-image. When the most popular images online are edited or completely artificial, it becomes harder to remember that real people do not look like that all the time.
At the same time, AI could also be used in a positive way. If developers train these systems with more diverse images, AI could help show a wider range of beauty. It could highlight different skin tones, ages, and facial features instead of repeating the same look over and over.
In the end, AI is not creating beauty standards from nothing. It is reflecting what already exists online. The real question is whether we want technology to keep repeating those standards or help expand them.
Sources:
- Bernal, I. C., Andre, J., Patel, M., & Newman, M. I. (2024). Beauty re-defined: A comparative analysis of artificial intelligence-generated ideals and traditional standards. Cureus.
- Dinkar, T., Jiang, A., Abercrombie, G., & Konstas, I. (2025). Erasing “ugly” from the internet: Propagation of the beauty myth in text-image models. arXiv.
- Makhseed, A., Arian, H., & Shuaib, A. (2025). Algorithmic bias in AI-guided aesthetic standards. Clinical, Cosmetic and Investigational Dermatology.
