Chain-of-Thought (CoT)
A prompting technique that instructs the AI to reason step-by-step before giving a final answer. Dramatically improves accuracy on complex reasoning, math, and logic tasks.
Why it matters
For math, logic, and multi-step problems, asking an AI to think out loud genuinely improves accuracy. It matters because a model rushed to a one-word answer often gets it wrong, while the same model reasoning through steps catches its own mistakes. It also makes the answer auditable: you can see the reasoning and spot exactly where it went off track instead of guessing.
A concrete example
Instead of "How many hours between 9:45am and 6:15pm?", add "Think step by step, then give the final answer." The AI lays out the arithmetic, minutes then hours, and lands on 8.5 hours more reliably. For anything with several moving parts, a word problem, a scheduling puzzle, a comparison, that simple phrase noticeably cuts down careless errors and shows you its work.
How to use it
Ask for reasoning on anything with steps: arithmetic, logic, planning, multi-part comparisons, or deciding between options against several criteria. "Work through this step by step before giving your answer" measurably improves that class of task. It also makes errors findable, because you can read the reasoning and see exactly where it went wrong instead of just disagreeing with a conclusion. The effect was documented in Chain-of-Thought Prompting Elicits Reasoning in Large Language Models, which is worth reading for how small the change is.
The common mistake
Trusting the stated reasoning as an account of how the answer was actually produced. It is generated text that tends to correlate with better answers, not a transcript of the model's internal process — reasoning can look sound and support a wrong conclusion.
Related terms
Prompt Engineering
The skill of crafting effective AI prompts to achieve desired outputs. Techniques include role-setting, few-shot examples, chain-of-thought reasoning, and constraint specification.
AI Agent
An autonomous AI system that can perceive its environment, make decisions, and take actions to achieve goals — like managing your email, scheduling meetings, or monitoring data.
Benchmark
A standardized test or dataset used to evaluate and compare AI model performance. Common benchmarks include MMLU (knowledge), HumanEval (coding), and MT-Bench (conversation). Helps users choose the right model for their needs.
A/B Testing (for Prompts)
Running two or more prompt variations against the same AI model to compare output quality. Data-driven prompt optimization that consistently outperforms intuition-based prompting.
Few-Shot Prompting
Providing 2-5 examples of desired input-output pairs in your prompt before asking the AI to perform the task. Significantly improves output quality for specialized tasks.
System Prompt
A special instruction set given to an AI model before the user's message, defining the model's persona, behavior rules, output format, and constraints. The foundation of custom AI assistants and chatbots.