EXPLAINER: Ever wonder why AI people somehow look like they’re all related?

AI repeatedly encounters certain ideas of what a “good-looking” human being is supposed to look like.

You’re scrolling through Facebook when another obviously AI-generated photo appears.

Maybe it’s a beautiful woman selling a product. A handsome young man supposedly working on a farm. A Filipino family standing inside a suspiciously spotless home.

Different characters, different situations—but somehow, you’ve seen those faces before.

The smooth skin. Perfect teeth. Symmetrical features. Sharp jawlines. Cinematic lighting. And that unmistakable glossy complexion that sits somewhere between human skin and expensive plastic.

Why do AI people seem to all look alike?

Part of the answer lies in what artificial intelligence learned from us.

Image generators are trained on enormous collections of existing images. And the internet itself isn’t a neutral portrait of humanity. It is packed with advertising, stock photography, celebrity photos, professionally lit portraits and heavily edited social media images.

That means AI repeatedly encounters certain ideas of what a “good-looking” human being is supposed to look like: clear skin, balanced features, prominent cheekbones and conventionally attractive proportions.

Then comes averaging.

Generative AI works through patterns and probabilities. Ask it simply for “a beautiful woman” or “a handsome man,” and without enough specific instructions, it can gravitate toward highly represented visual patterns in its training and tuning.

The strange mole, uneven eyebrows, acne scars, tired eyes, crooked teeth, sun damage and countless tiny asymmetries that make real people recognizable may disappear.

What remains can feel like humanity after a beauty filter.

There’s another reason: many AI products are built on related underlying models or tuned toward similar popular aesthetics. Some systems may also expand or reinterpret simple prompts to produce more visually appealing results.

So even when users type different instructions, the finished images can converge on the same visual vocabulary: dramatic lighting, shallow depth of field, flawless faces and hyper-detailed skin.

It explains why spotting AI images can sometimes feel strangely easy.

It isn’t necessarily the six fingers anymore.

It’s the face.

The irony is that AI can theoretically create almost any face imaginable. But when we give it vague instructions, it often retreats toward familiar patterns.

Want a more believable AI character? Specificity matters. Give the person an exact age, facial asymmetry, pores, wrinkles, scars, imperfect teeth, ordinary clothes and unflattering or natural lighting. Describe a candid moment instead of asking merely for something “photorealistic” or “beautiful.”

Because real people aren’t averages.

Perhaps that is exactly why, when AI tries too hard to create the perfect human face, it sometimes creates the same human over and over again.

READ: