Skip to content
AI · MEDIA · ONLINE DECEPTIONEvidence before belief
← All examples

AI images · Worked scenario

A convincing image. No original source.

Before deciding an image is AI, find out what its appearance can—and cannot—establish.

Invented example for learning

A photograph-like image of a flooded railway station arrives in a group chat. One person says it proves a disaster happened today; another says the unusually smooth faces prove it was generated by AI. Nobody supplies an original link.

What is actually known?

You have an image and two competing claims. The forwarding chain does not establish who made it, when it was made or what it depicts. Both the disaster claim and the AI accusation need their own evidence.

How to investigate it

  1. Ask for the original post or file, then record the uploader, date and exact claim.
  2. Look for earlier appearances of the same image. Compare the whole frame, not just a visually similar scene.
  3. Check the location and claimed event against original reporting or records from the people responsible for the place.
  4. Inspect available metadata or Content Credentials. Missing metadata is a gap, not evidence of AI generation.
  5. Treat visual oddities and detector outputs as leads. Check whether compression, editing or an ordinary photographic effect could also explain them.

Where the evidence stops

The image's appearance alone does not establish either a real disaster or AI generation. A creator's attributable explanation, original production record or corroborated provenance could change that assessment. Until then, leave the origin unresolved.

This is an illustrative scenario, not a report about a real incident or a published verdict. No media, source record or detector result has been invented as evidence.

Take the next step

How to check an image for AI clues →

Check something of your own ↗Our methodology