The Turing test asks one question: can a machine's responses fool a person into thinking they're talking to another person? For decades it's been treated as the benchmark for "real" AI. For a support chatbot, it's not just the wrong test — chasing it actively makes the product worse.
Why fooling anyone was never the goal
A visitor asking a support chatbot about your refund policy doesn't care whether the thing answering them is human. They care whether the answer is correct. Optimizing for "sounds convincingly human" spends effort on something customers never asked for, and — worse — a chatbot that's good at seeming certain and personable is also good at seeming certain and personable while being wrong. 5 things everyone gets wrong about AI chatbots for support covers this exact myth: customers forgive a chatbot for being a chatbot. They don't forgive a confident wrong answer, human-sounding or not.
There's also an honesty problem
Deliberately trying to pass as human sits uncomfortably close to deception. A support interaction works better, not worse, when a visitor knows they're talking to an AI — it sets the right expectations for what it can and can't do, and makes an eventual "let me get you a human" handoff feel like the system working as intended, not a bait and switch.
A better test: would you act on this without checking?
Here's a more useful question than "could this fool someone": if a customer made a real decision — canceled a plan, changed a setting, told their manager something was true — based purely on what the chatbot just told them, would that have gone fine? Call it the trust test. It has nothing to do with tone or personality and everything to do with whether the answer was actually grounded in something real. This is the same standard behind the case for boring AI: reliable and slightly robotic beats charming and occasionally wrong, every time real money or a real decision is on the line.
What passes the trust test in practice
Not a chatbot that never sounds like a machine — one that's right when it answers and honest when it can't be, the behavior described in grounded vs. hallucinated answers. Nobody needs to be fooled by that. They just need to be able to rely on it.
Related reading
- 5 things everyone gets wrong about AI chatbots for support — including the myth this post is arguing against directly.
- The case for boring AI — why unglamorous and reliable beats impressive and human-sounding.
- Grounded vs. hallucinated answers — what actually determines whether an answer passes the trust test.