What is Chain of Thought? — AI Glossary

What it is: Chain of Thought (CoT) is a technique for getting better answers from AI models by asking them to “think step by step” before answering. Modern reasoning models do this automatically; older models do it when you ask them to.
Who it is for: Anyone using AI for problems that need real reasoning — math, logic, multi-step planning, complex analysis.
Best if: Your AI answers feel shallow or jump to wrong conclusions. Asking for chain-of-thought often fixes it.
Skip if: You’re asking simple factual questions where extra reasoning doesn’t help (e.g., “capital of France?”). Want one practical AI workflow every morning? Subscribe to our free daily newsletter.

What is chain of thought?

Chain of Thought (often shortened to CoT) is a prompting technique where you ask an AI model to show its reasoning step by step before giving a final answer. Instead of jumping straight to a conclusion, the model walks through intermediate steps — and that walk dramatically improves accuracy on tasks involving math, logic, multi-step planning, and complex reasoning.

The technique was formalized in a 2022 Google paper showing that adding “Let’s think step by step” to a math problem could double or triple a model’s accuracy. Since then, CoT has become standard practice — and the newest generation of reasoning models (OpenAI’s o-series, Claude’s extended thinking mode, Gemini Deep Think) do it automatically without being asked.

Why does chain of thought matter?

Chain of thought is the single most useful prompting technique a beginner can learn. It costs almost nothing to add to your prompts, and the accuracy improvement on hard problems is large enough that it changes which problems are tractable for AI.

CoT also revealed something important about how language models work. The same model that fails at a tricky math problem on first try often gets it right when forced to think through intermediate steps. That suggests the “intelligence” is there but needs to be unlocked by the prompt — a finding that shaped the design of every reasoning model that followed.

How do you use chain of thought in practice?

The simplest version: append “Let’s think step by step” or “Show your reasoning” to your prompt. The model will lay out its thinking before answering. You can then review the steps and catch errors.

More advanced versions:

• Few-shot CoT — show the model 2-3 examples of solving similar problems step-by-step, then ask it to solve yours.
• Self-consistency — ask the model to solve the same problem multiple times with different reasoning paths and take the majority answer.
• Tree of Thoughts — a fancier version (see Tree of Thoughts) where the model explores multiple branches and prunes weak ones.

For reasoning models like o1, Claude with extended thinking, or Gemini Deep Think — you don’t need to ask for CoT. They do it automatically and bill you for the “thinking tokens” consumed.

Related terms

• Prompt Engineering
• Tree of Thoughts
• Reasoning Models
• Large Language Model

Learn more on Beginners in AI

• Best Claude Prompts
• AI Glossary

Sources and further reading

• Chain-of-Thought Prompting Elicits Reasoning (Google, 2022)

Last reviewed: May 2026. AI terminology evolves quickly — verify specifics on the official source pages above.

You may also like

• Best Claude Prompts
• AI Glossary
• Prompt Engineering
• Tree of Thoughts
• Reasoning Models
• Large Language Model