Chain of thought refers to intermediate reasoning steps that may help explain or structure how a problem is approached, especially in language-model prompting and evaluation discussions.
The term should not imply that AI systems must expose hidden private reasoning. In professional use, teams often prefer concise reasoning summaries, verifiable evidence, and final answers over raw model traces.