Can you catch it?

What the Catch Test teaches

Compare an AI answer with the information it was given before you act. Practice spotting a missing condition, an unsupported assumption or a wrong calculation. Accept a sound answer when the evidence supports it.

A synthetic worked example

This is a fictional practice case from the game, not a customer result or a captured model response. A travel receipt lists train tickets at $126, a taxi at $42 and a booking fee at $12. Three people share the cost equally. The practice AI answer says: “The total is $180, including the booking fee. Split three ways, that's $60 each.”

Check it: $126 + $42 + $12 = $180. Then $180 ÷ 3 = $60. Choose “It checks out. $60 each is right.” The evidence supports this answer. The useful habit is checking, not assuming every AI answer is wrong.

How scoring works

Each of the six rounds has one keyed answer. Each matching choice earns one point, for a total from zero to six. Five rounds contain a problem; one contains a sound answer. The result also reports problems caught out of five and false alarms on the one sound answer. There is no speed score.

What the score cannot tell you

These fixed, fictional cases are teaching examples. They are not a validated assessment of general ability, a representative survey, a hiring test or a model benchmark. Repeating the same cases can reflect memory. No live model is being tested here.

Practice these checks with your team in AI Lab, try a free workbook preview, or read the AI Readiness Assessment scope.