What Is the Turing Test? A Plain-English Guide

Whenever a new chatbot goes viral, someone asks whether it has passed the Turing test. It's a famous phrase, and it gets used in headlines as if it were an official certificate. It isn't. I'll explain what the test really is, where it came from, and why it matters less than people think when you're choosing AI tools for your business.

Where the Turing test comes from

Alan Turing described the idea in his 1950 paper, "Computing Machinery and Intelligence". He called it the imitation game. According to Wikipedia, the original game was a party game for three people, where one person questioned two others in another room and tried to work out which was which.

Turing's clever move was to set aside the huge question "Can machines think?" and ask something you can check instead: could a digital computer do well in that imitation game? The Stanford Encyclopedia of Philosophy describes him as treating the first question as too vague to discuss, which is why he swapped it for a game with clear rules.

How the Turing test works, step by step

The modern version is simple to picture. A human judge has a text conversation with two unseen parties, one a person and one a machine, and then has to decide which is which.

1. A judge types questions to two hidden partners, with no voices and no faces.
2. One partner is a person who answers honestly. The other is a machine trying to seem human.
3. After the conversation, the judge decides which partner is the machine.
4. If the judge can't reliably tell, the machine is said to have passed.

Notice what's being measured. Wikipedia points out that the result depends on how closely the answers resemble a human's, not on whether those answers are correct. A machine that makes small typing slips and cracks a joke might do better than one that is flawlessly accurate.

Has any AI passed the Turing test?

You'll find confident claims both ways, so here's how I read it. The Stanford Encyclopedia of Philosophy notes that in 2014 a chatbot called Eugene Goostman was said to have passed after fooling 33% of judges, but the trial was small and the result couldn't be relied on to predict anything wider.

It also stresses that a serious test means surviving repeated questioning from someone who knows a machine might be on the other end, rather than fooling casual observers. So when a headline says a chatbot has passed, check three things: how many judges, how long the chats lasted, and whether the judges knew what they were looking for.

Why many experts say the Turing test falls short

The test has been argued over for decades. The best-known objection is the Chinese room argument from John Searle in 1980. Wikipedia sums it up as the claim that passing can't prove a machine is thinking, because a program might shuffle symbols convincingly without understanding any of them.

Here are the other limits worth knowing:

• It rewards sounding human. Being helpful and being human-like are different goals.
• It only tests conversation. Nothing about maths, planning or doing real-world tasks.
• It depends on the judge. A sceptical, well-prepared judge is much harder to fool than a distracted one.
• It doesn't measure honesty. A machine can pass by bluffing.

The Stanford entry adds a fair point in Turing's defence: his claim was narrow. If something can hold a conversation with us, we have good grounds to credit it with intelligence. He didn't say only conversation counts.

What the Turing test means when you use ChatGPT, Claude or Gemini

Today's assistants write fluent, friendly text, so many people feel they'd easily fool a casual judge. That's exactly why the test is a poor shopping guide. Fluency tells you a tool sounds convincing. It doesn't tell you it's right.

Treat the sentence "it sounds just like a person" as a warning rather than a guarantee. If you want to understand why these tools can sound sure of themselves, my explainer on where ChatGPT gets its information walks through what's going on under the surface.

Better ways to judge an AI tool for your business

Instead of asking whether a tool could pass for a human, try questions that matter to your work:

• Does it follow my instructions the same way each time?
• Are the facts it gives me correct when I check them?
• Does it keep my tone and my customers' details safe?
• Does it save me time on a task I do every week?

A short trial on a real task beats any famous test. If you'd like ideas for tasks to try, start with ways to use AI in everyday life, and if you'd rather have someone help you choose and set things up, that's what I do as an AI consultant.

Turing test mix-ups to avoid

• It isn't a law or a standard. Nobody awards a pass certificate.
• It isn't the same as artificial general intelligence. Wikipedia lists it as one idea people have suggested for spotting AGI, not a proof of it.
• It isn't the CAPTCHA you click on websites. Those puzzles are a reverse check for spotting bots, and a different idea.

Frequently asked questions

What is the Turing test in simple terms?

It's a test proposed by Alan Turing in 1950. A judge chats by text with a person and a machine without knowing which is which. If the judge can't reliably tell them apart, the machine passes.

Who invented the Turing test?

Alan Turing, in his 1950 paper "Computing Machinery and Intelligence", where he called it the imitation game.

Has ChatGPT passed the Turing test?

There's no official pass. Researchers have reported results for some systems, but they depend on how the test is run, and experts still argue about what a pass would show.

Does passing the Turing test mean an AI can think?

Not necessarily. Critics such as John Searle argue a program could imitate conversation without understanding it, so the test shows imitation rather than thinking.

Is the Turing test still used today?

It's mostly a talking point and a teaching tool. Researchers use many other tests that measure specific skills, and businesses are better off judging a tool on their own tasks.