Chain of Thought is a prompting technique where a model is encouraged to reason step by step before giving its final answer, which improves its accuracy on complex problems.
What it means in plain English
Asked a hard question, a model that jumps straight to an answer often gets it wrong. Chain-of-thought prompting — for example, adding “think step by step” — encourages the model to work through the problem in stages, showing its reasoning. Laying out the intermediate steps helps it reach a correct conclusion, much as a person is more accurate when they work through a problem on paper rather than guessing.
It is especially effective for maths, logic, and multi-step reasoning tasks.
A simple example
For a word problem involving several calculations, simply asking for the answer may produce a mistake. Adding “solve this step by step, showing your work” prompts the model to reason through each stage, and it is far more likely to get the right final number.
Why it matters
Chain-of-thought is one of the most important discoveries in prompting. It reliably boosts performance on reasoning tasks, and the idea of getting models to “think” before answering has become central to how the most capable AI systems are built and used.
Related terms
- Prompt Engineering — the skill chain-of-thought is part of.
- Large Language Model — what benefits from step-by-step reasoning.
- Few-Shot Learning — often combined with chain-of-thought.
Frequently asked questions
What is chain-of-thought prompting?
It is prompting a model to reason step by step before giving an answer, which often improves accuracy on tasks that require multi-step reasoning, like maths or logic problems.
Why does step-by-step reasoning help?
Breaking a problem into intermediate steps lets the model work through it more carefully, rather than jumping straight to an answer, reducing errors on complex tasks.
