What is chain-of-thought prompting?

Chain-of-thought prompting is asking the model to show its work — to reason step by step out loud — before it gives you a final answer, instead of jumping straight to a conclusion. In practice, it's as simple as adding a phrase like "think step by step" or "explain your reasoning before answering" to your prompt.

Why does chain-of-thought prompting work?

On a multi-step problem — a math word problem, a logic puzzle, a question with several parts — a model that jumps straight to an answer can quietly take a wrong turn early on, and you'd never see it happen. Forcing it to write out the intermediate steps is like asking a person to show their work instead of blurting a guess: errors that would happen silently in one leap become visible, and often get caught and corrected, when they're spelled out one step at a time. The reasoning it generates also becomes part of what it reads next, so each step builds on a stated foundation rather than a hidden hunch.

When should you use it?

It pays off most when there's actually something to work through, and adds little — sometimes just noise and cost — when there isn't.

• Great fit — math, logic puzzles, multi-part questions, planning, and anything needing several steps of inference.
• Poor fit — simple factual lookups or one-word answers, where there are no intermediate steps to expose.
• Bonus — the visible reasoning doubles as an explanation you can check, so you're not trusting a black-box answer.

How is it different from a reasoning model?

Chain-of-thought prompting is the manual, DIY version. Reasoning models do the same kind of step-by-step thinking automatically and internally, by design, without you having to ask for it. Prompting it yourself is a widely used trick for getting better answers out of an ordinary model, and it still works today — it costs you nothing but a few extra words in your prompt.

With a dedicated reasoning model the behavior is built in, so you get it whether or not you spell it out.