Judge the trail, not the picture
How AI pictures are made
Many image tools start from random-looking noise and refine it step by step, guided by your words and by visual patterns learned in training, until a picture appears. Not every tool works the same way, but the idea holds: the picture is newly generated, not found. The same tools can also edit a real photo, changing one object, a face or the background while keeping the rest.
Why spotting fakes by eye fails
You may have heard to look for extra fingers, garbled text or odd shadows. Don't rely on it. Current tools often leave no visible flaw, and genuine photos can look strange for innocent reasons such as lens distortion or compression. A weird detail is a reason to investigate, never proof either way. Detector websites are also unreliable, so do not treat their percentages as a verdict.
Four questions that do work
Who is the original source, and can you find them? Is there a label or credit, such as a news agency or a Content Credentials mark? When was it first posted, and does that fit the story? Does the context fit: the place, the weather, other coverage of the same event? A reverse image search often answers the first and third questions in under a minute.
Not enough information is an answer
Often the honest conclusion is "I can't tell yet", and that is fine. The useful step is not to share until a trusted source confirms it. It also helps to remember the upside: the same tools power film effects, art, restoring old family photos and accessibility features. Generation is not the problem; passing something off as what it isn't is.
Have you shared a picture or clip recently without knowing where it first came from?