Make AI show its reasoning chain-of-thought prompting
Ask the model to work through it step by step, and you can catch where it goes wrong.
In short
Chain-of-thought prompting means asking the AI to work through its reasoning step by step before giving the answer.
- It improves accuracy on anything with logic: math, multi-step decisions, comparisons, word problems.
- The reasoning is also a check: you can see where the model went wrong instead of only trusting the result.
- Newer reasoning models do this on their own; for the rest, add 'think step by step' or 'show your work.'
Why showing the work helps
Chain-of-thought prompting is asking the model to reason out loud before it answers, instead of jumping straight to a result. Given a multi-step problem cold, a model often blurts a confident answer that skipped a step and landed wrong. Told to work through it step by step, it slows down and gets it right more often. You also get a visible trail you can check. The newest reasoning models do this automatically; with everyday models you switch it on with a line. When a task needs more than one pass of reasoning, prompt chaining splits it properly, and visible reasoning is also one way to reduce AI hallucinations. Both sit inside prompt engineering.
How to use chain-of-thought prompting
Turn on the reasoning, then read it as a check on the answer.
- 1
Add the trigger phrase
Append 'think step by step' or 'show your reasoning before the answer.' That alone switches on the slower, more careful mode.
- 2
Ask for the steps, then the answer
'Work through this step by step, then give your final answer on the last line.' You get the reasoning and a clean result.
- 3
Read the steps, not only the result
Scan the reasoning for the step that looks off. That is where errors hide, and they are easy to spot once written out.
- 4
Point at the bad step and rerun
If a step is wrong, say 'step 3 is off because...' and let it redo from there instead of starting over.
- 5
For hard problems, use a reasoning model
GPT-5, Claude extended thinking, and similar modes reason by default. Reach for them when the logic is genuinely complex.
Watch-outs
- Chain-of-thought is for reasoning tasks. For a simple lookup or a quick rewrite, it just adds noise.
- A convincing chain can still be wrong. Check the steps, especially any arithmetic.
- Longer reasoning costs more time and, on some tools, money. Use it where accuracy matters.
Common questions
What is chain-of-thought prompting?
It is prompting the AI to spell out its reasoning step by step before giving an answer, rather than jumping to a result. Working through the steps improves accuracy on logic-heavy tasks and gives you a visible trail you can check for mistakes.
Does 'think step by step' actually work?
Yes, on tasks that involve reasoning: math, multi-step decisions, and comparisons. Forcing the model to lay out the steps catches errors it would otherwise skip past. It does little for simple factual lookups, where there is nothing to reason through.
When should I use chain-of-thought prompting?
When a problem has more than one step or needs logic to get right: calculations, planning, weighing options, or anything where a skipped step changes the answer. Skip it for quick rewrites and simple facts.
Do I still need it with reasoning models?
Less often. Reasoning models like GPT-5 and Claude extended thinking work through steps automatically. The trigger phrase is mainly for standard models, and you can still ask any model to show its reasoning so you can verify it.
See how you actually work with AI
The AI Skills Quiz scores your real habits, including whether you check the model's reasoning or take the answer on faith. Free, and it takes a couple of minutes.
Professionals from these companies build their AI skills with Candova
The future belongs to AI‑native companies.
Make sure yours is one.
We make your people AI-fluent so your company becomes AI-native, with Cando, a personal AI trainer for every employee. Start with Candova AI today.
Free to start · Productivity gain guarantee · Built for your real job