Definition
A 1950 thought experiment: a machine passes if a judge chatting by text cannot tell it apart from a real person.
At a glance
- Proposed by Alan Turing in 1950, originally called the imitation game.
- A text-only behavior test: pass if a judge can’t reliably tell the machine from a human.
- In a 2025 UC San Diego study, GPT-4.5 was judged human 73 percent of the time.
- Passing means convincing imitation, not real understanding or truthfulness.
How it works
A judge types back and forth with two hidden partners, one human and one computer, and guesses which is which[1]. If they can’t reliably tell them apart, the machine passes[2]. It’s text only, so looks and voice don’t count.
Has anything passed it
For decades, nothing did. Then GPT-4 was judged human about 54 percent of the time in 2024, and GPT-4.5 with a persona hit 73 percent in 2025, often beating the real humans[3]. By Turing’s original yardstick, modern AI now passes[4].
Why it matters for your business
Customers increasingly can’t tell your chatbot from a person. That makes AI support cheaper and more natural, but it can still state errors confidently, so honesty and trust matter. Many regions and companies now disclose when a customer is talking to a bot[5].
Bottom line
AI now clears the conversation bar, so the real question is whether and when you should tell customers your chatbot isn’t human.