7.3 Search, Chat, and Reasoning

Everyday AI experiences can be understood by their main job, without memorizing product lists. A search-oriented experience focuses on finding current pages or information and connecting answers to sources. A chat-oriented experience focuses on following instructions, explaining, rewriting, and continuing a conversation. A reasoning-oriented experience spends more effort handling constraints, comparing paths, or working through a complex problem.

These are not sealed boxes. One product may support all three experiences. A model that normally answers from available context may search first and synthesize later when a web tool is enabled.

Choose a starting point from the task

For today's opening hours, a recent policy, or a current price, prefer a search experience that shows current sources—and open the original pages. For explaining a concept, rewriting an email, or brainstorming, a chat experience is often natural. For a task with many constraints and paths to compare, deeper reasoning may be a better fit.

“Reasoning longer” does not guarantee correctness, and every simple task does not need the slowest mode. A complicated but mistaken process can still reach a wrong conclusion. In every mode, provide clear input and verify important facts using the methods from Chapter 6.