
Odd details can be a clue: weird text, odd hands or teeth, lighting that doesn’t add up, or a pattern that repeats oddly. A clean picture can still be AI.
Check the original file
Don’t trust a screenshot. Hidden labels often live in the original file. Some AI images carry a hidden label (Content Credentials or SynthID).
- openai.com/verify finds OpenAI’s labels.
- For Google AI: upload the original file to Gemini and ask, “Was this created or edited by Google AI?”
A “no label” result is not proof it’s a real photo. Labels can be removed, and other tools may leave none you can check.
If it involves news or money
Pause. Do a reverse image search, check a trusted news site, and if someone rushes you, wait.
Voice can be faked too. If a call asks you to send money, see That call from Mom might be AI.