Lateral Thinking Test
What is a lateral thinking test?
A lateral thinking test is a set of questions where the obvious answer is the wrong one. Each item is built around an assumption that most people make without noticing, and the correct answer comes from spotting that assumption rather than from working faster. The phrase was coined by Edward de Bono in 1967 to describe solving problems by approaching them from an unexpected angle, and puzzles of this kind have been a staple of interviews, pub quizzes and psychology lectures ever since.
This free version has 16 multiple-choice questions and a 12-minute limit. It reports a percentage with a breakdown across four categories, and because the same questions are put to AI models through the site's API, it doubles as one of the clearest demonstrations of where people still out-think machines. You can compare your result on the human-versus-AI leaderboard.
The four categories
- Misleading maths — arithmetic that looks trivial and is designed to make you answer before you have read it. The bat-and-ball problem is the classic: a bat and ball cost £1.10 and the bat costs £1 more than the ball, so the ball is 5p, not 10p.
- Hidden assumptions — situations where the question quietly relies on something you have assumed but were never told, such as the sex of a doctor or the number of people in a room.
- Word traps — questions that turn on the exact meaning of a word, a plural, or a phrase that is usually read loosely.
- Everyday logic — ordinary scenarios where the sensible answer is not the one the wording points you towards.
How the test is scored
Each question has one correct option and a short explanation, shown after the test, of why the tempting answer is wrong. Your headline score is the percentage correct out of 16, with the four category scores alongside it. Answer options are shuffled every time, so retaking the test does not reward memorising letter positions. The percentile shown compares you with other players on this site rather than with a population norm, because lateral thinking is not a standardised construct with published norms.
Why trick questions defeat AI
A language model produces the most probable continuation of a piece of text. A trick question is engineered so that the most probable continuation is wrong. Worse, the famous puzzles appear thousands of times in training data with their famous answers attached, so a model that has learned "bat and ball → 5p" may confidently answer 5p to a variant whose numbers have been changed. People make the same error when they skim, which is why the test rewards the same thing in humans and machines: reading the question that was asked rather than the one you expected.
The leaderboard makes the gap measurable. Alongside this test the arena includes the Common Sense Test, the Token Trap Test and the Visual Reasoning Test, each chosen because it probes a different weakness of text-trained models.
Tips for a higher score
- Read every question twice before you look at the options. Nearly every trap is in the wording, not the arithmetic.
- Ask what you are assuming. If a question feels easy, name the fact you are relying on and check whether it was actually stated.
- Distrust round numbers and familiar puzzles. If you recognise a classic, assume the numbers or the twist have been changed.
- Use the time. Twelve minutes for 16 questions is deliberately generous; the test measures care, not speed.
Frequently Asked Questions
What is lateral thinking?
Solving a problem by questioning the assumption everyone else has made, rather than by following the obvious line of reasoning faster. The term was coined by Edward de Bono in 1967. In a test it shows up as riddles and trick questions where the first answer that comes to mind is the one the question was built to provoke.
How is the Lateral Thinking Test scored?
There are 16 multiple-choice questions across four categories with a 12-minute limit. Your score is the percentage correct, with a per-category breakdown. Options are shuffled on every attempt.
Why do AI models fail lateral thinking questions?
Because the questions are anti-patterns. A language model predicts the most likely continuation, and a trick question is written so that the most likely continuation is wrong. Models trained on the famous versions often answer the famous version even when the wording has changed.
Is this a real psychometric test?
No. It is an original test for entertainment and personal development. Lateral thinking is not a standardised construct with published norms, so the percentile compares you with other players on this site only.
Can I improve at lateral thinking?
Yes, mainly by learning to pause. Most trick questions are beaten by re-reading the question once and asking which word you are assuming a meaning for. The transferable skill is the habit of checking the premise, not knowing particular puzzles.