Chain-of-Thought

Chain-of-thought is a prompting and model-design technique in which a large language model produces its intermediate reasoning steps before giving a final answer. Modern reasoning models (e.g., OpenAI o-series, Claude thinking modes) expose chain-of-thought as visible internal deliberation. For accessibility, chain-of-thought can improve trust: blind users of generative AI report that seeing the model's reasoning helps them confirm it interpreted the intent correctly and followed an appropriate process, especially for spreadsheet formulas and data analysis. However, reasoning controls are often poorly exposed to screen readers — placed between the prompt and the response — so users frequently miss them, and reasoning alone is never a sufficient substitute for checking the actual output.