GPT-4.5 Fooled 73% of Judges: 20 Turing Tests, Detectors and Fakes to Check Your AI Radar
People picked GPT-4.5 as human 73% of the time. Try 20 Turing tests, AI detectors and deepfake generators to see if you can still spot the AI.
In a 2025 study from UC San Diego, participants identified GPT-4.5 as human 73% of the time, more often than they recognized genuine humans. ELIZA, a chatbot from the 1960s, achieved 23%, closely followed by GPT-4o at 21%. The study comprised 20 Turing Test scenarios to help differentiate between humans and AI across chat, text, faces, art, voice, and video.
A three-party Turing test involving 284 participants revealed that GPT-4.5 with a persona prompt won 73% of the time, while LLaMa-3.1-405B secured 56%. The public versions of AI were found to be equally convincing, with participants correctly identifying bots only 68% of the time and humans 73% of the time. The test extended beyond chat, with a blind listening study showing listeners were essentially guessing at close to 50% accuracy when trying to distinguish AI-composed piano music.
Written by urgent.news from HackerNoon's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
