When you suspect an image might be AI-generated, what kinds of clues and checks help you decide?

Don't trust the surface — verify provenance and look for the tell-tale inconsistencies that generative models still produce.

Whether a photo of, say, a wrecked vehicle in a forest or a person on a street is real or AI-made often can't be judged by eye at a glance, so you combine looking with checking:

• Hunt for inconsistencies generative models still struggle with — malformed hands and text, impossible reflections or shadows, garbled background details, physically wrong geometry.
• Establish provenance — where did this image first appear, and does any reputable source actually report the event it claims to show? A dramatic image with no credible reporting behind it is a red flag.
• Check context — does the scene match the time, place and weather it's attributed to?

The mindset shift is that a realistic image is no longer self-validating; "it looks real" tells you nothing, so the question becomes "can I trace it to a trustworthy origin?"