Why does CoT improve AI reasoning?

Since LLMs generate one token at a time, CoT gives the model intermediate results to reference, preventing it from having to compress all reasoning into a single leap.

Which type of task benefits MOST from CoT?

CoT is most valuable for tasks requiring multiple reasoning steps. Simple recall and creative tasks don't benefit as much because they don't require intermediate logical steps.

What Is Chain-of-Thought Prompting?

Understand the technique that dramatically improves AI reasoning on complex problems.

7 min read

3 quiz questions

Chain-of-Thought (CoT) prompting is a technique where you ask the AI to show its reasoning process before arriving at a final answer. Instead of jumping straight to a conclusion, the model walks through intermediate steps — just like a student showing their work on a math exam.

This technique was first formalized by Google researchers in 2022, but the core idea is intuitive: when you force a model to reason through a problem step by step, it makes fewer errors. The intermediate steps act as a scaffold that keeps the model's "thinking" on track.

Large language models generate text one token at a time. Without CoT, the model must compress all its reasoning into a single jump from question to answer. With CoT, each intermediate step becomes part of the context for the next step, giving the model more "working memory" to solve complex problems.

Prompt

See the difference CoT makes on a reasoning problem.

Without CoT

Prompt: "A store has 15 apples. They sell 8, receive a shipment of 20, then sell 12. How many apples remain?" Answer: 13 apples.

With CoT

Prompt: "A store has 15 apples. They sell 8, receive a shipment of 20, then sell 12. How many apples remain? Think step by step." Step 1: Start with 15 apples. Step 2: Sell 8 → 15 - 8 = 7 apples. Step 3: Receive 20 → 7 + 20 = 27 apples. Step 4: Sell 12 → 27 - 12 = 15 apples. Answer: 15 apples remain.

Multi-step math and logic problems
Complex reasoning that requires combining multiple facts
Coding problems that involve algorithmic thinking
Analysis tasks where you need to weigh multiple factors
Word problems and puzzles with tricky setups

There are two main approaches: Few-Shot CoT, where you provide examples of step-by-step reasoning before your question, and Zero-Shot CoT, where you simply add a phrase like "Think step by step" or "Let's work through this." We'll cover Zero-Shot CoT in the next lesson.

Few-Shot CoT Example

Provides worked examples before your question to guide step-by-step reasoning.

Here's how to solve word problems step by step:

Q: Roger has 5 tennis balls. He buys 2 cans of 3 tennis balls each. How many does he have now?
A: Roger starts with 5 balls. He buys 2 cans × 3 balls = 6 balls. Total: 5 + 6 = 11 tennis balls.

Q: The cafeteria had 23 apples. They used 20 for lunch and bought 6 more. How many apples do they have?
A: Start with 23. Used 20 → 23 - 20 = 3. Bought 6 more → 3 + 6 = 9 apples.

Q: [YOUR PROBLEM HERE]
A:

CoT doesn't just improve accuracy — it also makes the model's reasoning transparent. If the final answer is wrong, you can inspect the intermediate steps to find exactly where the logic went off track.

Prompt Templates

Step-by-Step Reasoner

General-purpose CoT prompt for any reasoning problem.

I need you to solve this problem. Show your complete reasoning process step by step before giving the final answer.

Problem: [YOUR PROBLEM]

Walk through each step of your reasoning. Number each step. After all steps, clearly state your final answer.

Logic Problem Solver

Structured CoT for logic puzzles and deductive reasoning.

Solve this logic problem using careful step-by-step deduction:

[PROBLEM]

For each step:
1. State what information you're using
2. State what conclusion you can draw
3. Explain why that conclusion follows

After working through all deductions, state your final answer.