HomeCoursesStudyQuizzesPuzzlesFree TestsAI ArenaToolsDownloadsGuidesContact
Log InSign Up

Visual Reasoning Test

What is a visual reasoning test?

A visual reasoning test asks you to take information straight from a picture: the time on a clock face, the missing cell in a coloured grid, or which of four shapes is the original after it has been turned. There is nothing to read and nothing to calculate. That makes the questions quick for most people and surprisingly hard for artificial intelligence, which reasons mainly through language and has to translate a picture into words before it can think about it.

This free test has 12 questions in three categories and a 12-minute limit. Every image is generated by the site, so each puzzle has exactly one correct answer and no ambiguity in the drawing. Your percentage score and category breakdown appear instantly, and because the same puzzles are put to AI models through the site's API, you can see how your perception compares with theirs on the human-versus-AI leaderboard.

The three categories

  • Clock reading — tell the time from an analogue clock face with no numerals. Children learn this at about six; published evaluations in 2025 found that leading vision-language models misread most clock faces, often confusing the hour and minute hands.
  • Grid logic — a small grid of coloured cells follows a rule across rows or columns and one cell is missing. Spot the rule and pick the cell that completes it.
  • Mental rotation — one shape is shown, then four candidates; exactly one is the same shape rotated rather than mirrored or altered. Mental rotation is one of the best-studied spatial abilities in psychology, first measured by Shepard and Metzler in 1971.

How the test is scored

Each question has a single correct option. The headline score is the percentage correct out of 12, shown with your score in each of the three categories so that a clock slip and a rotation slip are reported separately. Answer options are shuffled every attempt. The percentile shown afterwards compares you with other players on this site rather than a population norm: this is an original test written for interest and self-reflection, not a validated psychometric instrument.

Why pictures still favour people

Modern AI systems can describe an image in fluent prose, which makes their failures at simple perception easy to miss. Describing is not the same as measuring. To read a clock the model must estimate two angles precisely; to complete a grid it must hold several cells in mind at once; to judge a rotation it must compare geometry rather than labels. Each step is natural for a visual system trained by a lifetime of looking at things and awkward for one that has mostly learned from captions. Scores are improving as models are trained on more diagrams and synthetic images, and the leaderboard records that progress.

The Visual Reasoning Test is one of four arena tests, alongside the Common Sense Test, the Lateral Thinking Test and the Token Trap Test. For a longer spatial workout, the Pattern Reasoning Test has 24 abstract matrix items.

Tips for a higher score

  • On clocks, find the short hand first. Its position between two hour marks also tells you roughly what the minute hand should show, which catches misreads.
  • On grids, check the rule across a full row and a full column before choosing; a rule that only works one way is a distractor.
  • On rotations, pick one feature of the original shape and follow it. A mirrored candidate will have that feature on the wrong side.
  • Do not rush. Twelve minutes for twelve images is ample, and every category rewards a second look.

Frequently Asked Questions

What does the Visual Reasoning Test measure?

Three visual skills: reading the time from an analogue clock face, completing the pattern in a small colour grid, and recognising a shape after it has been rotated. All three depend on perceiving a picture as a whole rather than describing it in words.

How is the Visual Reasoning Test scored?

There are 12 multiple-choice questions, four in each of three categories, with a 12-minute limit. Your score is the percentage correct, with a per-category breakdown. Answer options are shuffled on every attempt.

Why is reading a clock hard for AI?

Because the answer lives in the geometry of two hands, not in any text. A vision-language model has to estimate the angle of each hand and map both to hours and minutes, and published evaluations have found that even strong models misread the majority of analogue clock faces.

Is this a real psychometric test?

No. It is an original test for entertainment and personal development, built with generated images so each puzzle has a single correct answer. It has no published norms; the percentile compares you with other players on this site.

How do I get better at mental rotation?

Pick one distinctive feature of the shape and track only that feature through the rotation before checking the rest. Spatial games, jigsaws and sketching objects from different angles all transfer to this skill.