Study: The Real AI Deepfake Danger Isn't Fake Videos — It's Distrust

A researcher named Eilaf Mohamed studied how people in Sudan react to AI-generated fake videos and images. Her finding was surprising: there aren't that many AI deepfakes in Sudanese social feeds yet. Instead, the damage is coming from the other direction. People are accusing real videos of being fake, and they're increasingly unsure whether anything they see online can be trusted at all.

Why does that happen? Because when people decided whether a video was real or fake, they weren't using any real skill at spotting computer trickery. They were using what they already believed. If a clip supported their side, they accepted it. If it made their side look bad, they called it a deepfake. Researchers call this 'motivated reasoning' — believing what fits your existing views. That means the usual fixes, like better detection software, don't help much. Someone has to want to be corrected first.

The study warns this leaves countries dangerously exposed. Sudan is just the clearest example; the same weakness exists in the US, Europe, and anywhere with a heated political debate. When enough people stop expecting anything to be true, a condition the paper calls 'truth indifference' sets in. Once that habit takes hold, a future wave of convincing AI fakes would land easily, because nobody would trust the real evidence either.

The good news is that the recommended fixes are cheap, human, and don't require advanced technology. The paper suggests 'pre-bunking' — warning people about a lie before they see it, like a vaccine for misinformation — plus basic AI literacy lessons, fast fact-checking, and hotlines where people can ask whether a video is real. High-tech detection tools can come later, once institutions have the money and staff. The takeaway for anyone with a phone: skepticism alone isn't protection.