What Is Chain of Thought?
Also called chain-of-thought prompting, CoT
Short Answer
Chain of thought is a prompting technique where the model is asked to work through a problem in intermediate steps before giving the final answer. Showing the reasoning improves accuracy on math, logic and multi-step questions. Reasoning models now do this internally without being asked.
- Taught in lesson 7
Why It Matters
Asking a model to reason step by step can noticeably improve answers on problems with several steps, such as calculations, comparisons and planning. It also makes the model's reasoning visible, which helps people check its work.
How It Works
Google researchers described the technique in 2022 (Wei and colleagues), showing that adding worked examples with reasoning steps to a prompt sharply improved large models on math word problems. Even a simple instruction such as "think step by step" helps.
The steps a model writes are useful, but they are not a guaranteed true record of how it reached its answer. Check the final answer, not just whether the reasoning looks convincing.
There are two common forms. Zero-shot chain of thought simply asks the model to think step by step. Few-shot chain of thought shows examples that include worked reasoning. Reasoning models now do this kind of thinking internally, so explicit prompting matters less with them, but it still helps with standard models and when you want the steps shown. Visible reasoning is not always a faithful account of how the model reached its answer, so the final answer still needs checking.
Chain of Thought in Practice
Everyday Example
For "A shop sells pens at 3 for 2 euros; how much for 12 pens?", asking the model to compute the number of groups first leads to the right 8 euros.
Industry Example: Pharma
Asked to check whether a patient profile meets a trial's eligibility criteria, the model goes criterion by criterion before its verdict, which makes the reviewer's check faster.
Where You Meet It
- Prompts for maths, logic and planning tasks.
- Reasoning models that think before answering.
- Agents that plan their next steps.
- Explanations of how AI answers were reached.
Key Ideas Around It
- Step-by-step reasoning
- Breaking a problem into intermediate steps.
- Reasoning model
- A model trained to think before answering.
- Few-shot example
- A worked example included in the prompt.
- Self-consistency
- Generating several reasoning paths and choosing the most common answer.
A Common Mix-Up
Seeing reasoning steps does not guarantee a correct answer. Models can produce plausible steps that lead to a wrong result.
Check YourselfWhen does chain-of-thought prompting help most?
On problems with several steps, such as calculations, logic or planning, rather than simple lookups.
Learn It on ai2hired
S0 · #7Prompting from Zero: Get Better Answers
Coming in Season 0, Start Here. Every ai2hired lesson is free on YouTube.
Questions People Ask
Does chain of thought always help?
It helps most on multi-step problems. On simple lookups it adds tokens, cost and delay without much gain.
Do I still need it with reasoning models?
Usually not explicitly, because they reason on their own. Clear instructions and good examples still matter.
Related Terms
Learn AI From Zero to Hired
Free lessons on YouTube, each one built from what real AI job ads ask for.