AI Detectors are not Detectives

Where Human Fact-Checkers Will Always Make a Difference

Robert Bourgoing/28 September 2026

Upload an image, a video or a text to an AI detector and it may give you an answer in seconds: “85% likely to be AI-generated.”

It sounds precise. It is not proof.

AI detectors look for patterns and calculate probabilities. Text detectors such as Pangram, GPTZero, ZeroGPT, QuillBot and Copyleaks. Image detectors like Hive, Gemini’s SynthID, AI or Not, Sightengine and Illuminarty.

They can flag something suspicious, but they can also get things wrong. A real photo may be labelled fake. An AI-generated image may pass unnoticed.

That is why an AI detector is not an AI detective.

A detective does something different: checks the source, looks for inconsistencies, searches for earlier versions, compares evidence, verifies the context and asks whether several independent clues point in the same direction.

AI tools can help with that investigation. They can suggest leads, highlight anomalies and save time.

But they should never have the final word.

Use AI detectors for clues. Use your own judgment to verify the evidence.

Watch the short Terra Asia tutorial: AI Detectors Are Not AI Detectives