How to Tell if a Photo is AI Generated

Why it's getting harder

Three years ago, there was a reliable checklist. Hands with six fingers. Text in the background that was garbled. Teeth blending together into a smooth smear. Ears that looked like an afterthought. You could spot AI images pretty reliably just by looking carefully.

That checklist is still useful, but it's no longer sufficient. Current AI image generators — Midjourney v6, Stable Diffusion XL, DALL-E 3, and a dozen others — have gotten dramatically better at exactly the things that used to give them away. Hands are usually correct now. Text is often readable. Backgrounds are coherent.

The result is that visual inspection alone has become unreliable for many categories of AI-generated images, especially photorealistic portraits. Researchers at Northwestern University found that most people perform only slightly better than chance when asked to distinguish AI faces from real ones. The gap between what looks real and what is real has nearly closed.

That said, visual tells still exist. Knowing them helps you catch the ones that aren't perfectly rendered — and there are still plenty of those.

Visual signs a photo might be AI generated

These aren't universal — a clean result on all of them doesn't confirm a photo is real. But they're worth checking, especially if something already feels slightly off.

1. Hands. This was the classic tell, and it's still valid for lower-quality generations. Count the fingers. Look at the knuckle proportions. Hands gripping objects are particularly hard for AI to render correctly — the contact points between fingers and a cup or phone handle often look wrong.
2. Teeth. When the subject is smiling, look at the teeth. AI-generated teeth are sometimes too uniform — identical in size, too white, too perfectly aligned. Real smiles have variation. They also sometimes blend together at the edges or produce too many teeth for a normal mouth.
3. Ears. Human ears are complex shapes and AI models still struggle with them. Look at the inner ear structure. On a real photo, there's genuine three-dimensional detail. AI-generated ears often look smooth or slightly flattened, especially toward the back edge.
4. Hair near edges. Look at where hair meets the background. AI-generated hair often has an unnaturally smooth transition — the strands don't quite behave like they would in front of a real backdrop, especially when the background is complex.
5. Jewelry and accessories. Rings, earrings, necklaces. These tend to blur slightly into the skin or have an inconsistent metallic quality. AI generates accessories as part of the image rather than as separate objects, which shows at close range.
6. Background logic. Look at surfaces and objects behind the subject. Signs with readable text, bookshelves with identifiable titles, clocks with correct time — real photos have those. AI-generated backgrounds tend to have plausible-looking but incoherent details when examined closely.
7. Skin texture. Real skin, even well-lit skin, has pores, fine lines, and texture variation. AI-generated skin is often too smooth — not obviously fake, but unusually clean in a way that feels more like a rendering than a photograph. This is subtle and doesn't apply to all generators.
8. Eyes. The pupils should be consistent in size. The light reflections in both eyes should match each other. The iris texture should have natural variation. AI eyes are usually good now, but inconsistencies between the left and right eye still appear occasionally.

Again — a photo can pass all of these checks and still be AI-generated. These are the things that are still sometimes wrong. A detection tool checks for what you can't see: the statistical patterns in the pixel data itself.

What detection tools look for that you can't see

AI image generators produce images through a fundamentally different process than cameras. A camera captures light hitting a sensor. An AI model samples from a probability distribution learned from millions of training images. The output images look similar to photographs, but the underlying data has different statistical properties.

Detection models are trained to identify these differences. They look at patterns in the noise, color gradients, and pixel-level frequency distributions that differ between photographed images and generated ones. A clean, well-rendered AI image can be undetectable by eye but still identifiable by a model analyzing the raw image data.

This is why detection tools catch things visual inspection misses. The image looks real. But the math underneath it looks different from a real photograph.

The limitation of detection tools is that the generators keep improving. A detection model trained on 2023 outputs may not catch 2026 outputs as reliably. The best tools retrain on new generator outputs regularly, which is why accuracy figures need to be interpreted with the specific tool and timeframe in mind.

Types of AI-generated images and what each looks like

AI-generated portrait photos — photorealistic images of human faces. These are the ones used in fake social media profiles, catfishing, and fake reviews. They're the hardest to spot visually and the most important to check with a tool.

AI-generated scenes and environments — landscapes, city streets, indoor spaces without people. These often have the background logic issues described above. Signs, storefronts, and text on surfaces are usually the first place to look.

AI-generated art — illustrated, painterly, or fantasy-style images. These are less likely to deceive someone into thinking they're looking at a real photo, but they do matter in art theft and disclosure contexts. Artists and commissioners increasingly use detection tools to verify whether work was hand-made or generated.

AI-enhanced real photos — real photographs that have been heavily edited using AI tools. These are a separate category: the base image is real but has been significantly altered. Detection of this type is harder and less reliable than detecting fully generated images.

When visual inspection isn't enough — use a tool

Visual inspection is a first pass. For anything where accuracy matters — a dating profile you're about to invest time in, a news photo you're about to share, a job applicant's headshot — use a detection tool as the second step.

Faux Spy checks any image visible in Chrome. Right-click the photo (or hover for the Investigate button), click it, and you'll have a verdict with a confidence score in a few seconds. The five detection categories — AI Photo, Real Photo, AI Art, Digital Art, Illustrated — cover the main types of generated and real images you'll encounter.

For the specific use case of dating profile verification, the catfish detector guide covers the full workflow. For deepfake faces specifically, see the deepfake detector guide. For AI art on platforms like Etsy or DeviantArt, see the AI art detector page.