Turing Test
The 1950 proposal that a machine should count as thinking if it can pass for human in conversation — a test of deception, which Turing said plainly and everyone forgot.
Reviewed July 13, 2026Contested
When not to use it
- As a benchmark. No pass mark, no standard judge, no time limit. It's unfalsifiable in practice.
- As a definition of intelligence. Turing called it the imitation game and meant it — a behavioural substitute for a question he thought unanswerable.
- As evidence a system passed something meaningful. ELIZA passed with 200 lines of pattern-matching in 1966.
- To measure superhuman ability. A machine better than humans at arithmetic has to pretend to be worse to pass.
Reach for something else instead
- Capability evaluations — can it do this task, this reliably? No claim about minds.
- Winograd schemas — a serious attempt at world knowledge. Also solved.
- Task-specific benchmarks — narrow, measurable, honest about what they measure.
- Adversarial evaluation — test what breaks, not what convinces.
This entry is part of a longer guide: What is artificial intelligence?
Read more on the blog
- What AI cannot do, no matter how capable it getsFrontier models now rank just behind specialist verification tools at deciding whether programs terminate, which is supposedly undecidable. Both facts are true, and the reason they coexist is the thing most people get wrong.
- What is the Turing test? And why it stopped matteringTuring never proposed the imitation game as a definition of thinking. He proposed it to replace a question he considered meaningless. Machines have now passed versions of it, sometimes judged more human than actual humans, and the striking thing is how little that settled.
- Does it understand? The argument, properly statedBoth sides of this debate are usually presented by their opponents. Here is the sceptical case at full strength, the case for at full strength, why the two keep missing each other, and what would actually settle it.
Further reading
- Turing (1950), Computing Machinery and Intelligence — read it; it's short, funny, and much better than its reputation.
- Searle (1980), Minds, Brains, and Programs — the Chinese Room; forty-five years unresolved.
- Levesque, Davis & Morgenstern (2012), The Winograd Schema Challenge — the serious replacement, also solved.
Primary sources, listed so you can check the claims on this page rather than take them on trust.
Where people go wrong
- Citing it as the standard for machine intelligence. It's a proposal to stop asking that question.
- Treating a pass as significant. It's been passed; the field correctly ignored it.
- Forgetting it measures the judge. ELIZA's results were about people, not the program.
- Thinking the goalposts moved. They may have — or we may only ever learn what we meant by watching something achieve it.
At a glance
FieldFoundations
Year1950
Its real namethe imitation game
What it testswhether a machine can pass for human, which is deception, not thought
Statuspassed, and correctly ignored
The lessonevery test of "real intelligence" has fallen and been disowned
DifficultyBeginner
Often compared with
Turing Test vs. capability evaluation — one asks whether a machine can pass for human and gets you deception; the other asks whether it can do the task and gets you an answer.
Where this sits
A destination. 1 concept lead here, and nothing in the corpus depends on it.
1Levelsteps in
1Needs firstconcepts
0Opens upnothing further
1Areastays here
Learn these firstArtificial Intelligence
Computed from the prerequisite graph, not assigned. How this works