How to tell if an image is AI-generated
"Count the fingers" stopped working. Here is what still does.
Start with the source, not the pixels
- Who posted it first? A news agency, a known photographer, or an account made last week?
- Reverse image search (Google Lens, Bing, TinEye): does it appear anywhere earlier, or anywhere at all?
- Is there a caption, place and date someone could check?
- Content credentials: some cameras and editing tools attach a record of how an image was made. A "Content Credentials" or "cr" label shows it where supported.
Then look closely
- Text and signs: letters that melt into shapes, made-up words, logos that are nearly right.
- Light and shadows pointing different ways, reflections that don't match.
- Too perfect: glossy skin, cinematic light, everyone photogenic, background people with odd faces.
- Things that merge: earrings into hair, fingers into cups, railings that don't connect.
- Repeated patterns in crowds, leaves or bricks.
Good AI images can pass all of these. If it matters, trust the source check over your eye.
It's the same skill as text
Real things are messy and specific. Generated things are plausible and smooth. That's true of photos, and it's true of answers in a chat, which is what the daily game trains.
Questions
Do AI image detectors work?
They can help, but they make mistakes in both directions, especially after an image is cropped, compressed or screenshotted.
Are AI images labelled on social media?
Some platforms add labels when they detect them or when the creator declares it, but many AI images go unlabelled.