What if the old advice “solve it step by step” was the key to making AI reason better too? Just as people often crack tough problems by writing down each step, large language models can also improve when they are guided to “think out loud” before giving an answer.

Chain-of-thought (CoT) prompting is a technique where the prompt explicitly asks the model to reason through a problem step by step before producing a final response. It has proven especially useful for tasks that require reasoning, such as maths questions, logic puzzles, and common-sense problems. CoT gained prominence after a 2022 paper, “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models,” showed that giving a few examples with both the question and a worked step-by-step solution can dramatically improve performance. This approach also makes a model’s decision process more interpretable, since you see the intermediate steps rather than just the conclusion.

The next time an AI answer feels sloppy, try showing it a clear example and ask it to “show its working” or “reason step by step.” Getting AI to think out loud has pushed both researchers and everyday users to rethink how these systems reason and how careful prompting can bring out their best.