L

Learn AI

Track progress · learn offline

Open

Lesson 3 of 8

Give it room to reason

6 min read

Ask a quick logic or maths question and the AI often fires back a confident answer that's just wrong. Add four words — "think step by step" — and the same model gets it right. Those four words buy it room to work.

Blurting versus working it out

Left to answer instantly, an AI jumps to a guess — and on anything with a few moving parts, the fast guess slips. Ask it to lay out the steps first and it reasons its way there, catching the mistakes a snap answer makes. This is [chain-of-thought](glossary://chain-of-thought) prompting, and it's the same as a person: made to answer on the spot you might fumble the maths; given a moment to work it on paper, you get it right.

Asking the AI to reason step by step before answering — chain-of-thought — makes it far more reliable on anything multi-step, and lets you see the working instead of just trusting the answer.

When it helps, when it hurts

This isn't for everything. On a simple lookup or a bit of creative writing, "think step by step" just adds noise — there's nothing to reason about. Save it for questions with real steps: maths, logic, planning, multi-part decisions. And a caveat: neat-looking steps can still reach a wrong answer, so reasoning makes it more reliable, not infallible.

Reach for step-by-step on multi-step problems — maths, logic, planning. Skip it for lookups and creative writing, where it only clutters.

Careful-looking reasoning can still be wrong — the AI can lay out tidy steps and land on a bad answer. On anything that matters, check the result, not just that the working looks sound.

The gist

Which of these most deserves a "think step by step"?

Continue in the app

Take the whole Prompt Engineering course — tracked

Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.