An AI footprint lets users trace the steps between a prompt and an AI result. If a step looks mistaken, the user can identify it in a follow-up prompt and ask for a corrected result. The point is not transparency for its own sake: the trace should help someone judge and act on the output. section 10.2.2
In a conversational interface, explanations can appear within the conversation. When the process is largely opaque, as with a single-step LLM generation, links to sources can offer a more useful shortcut than a step-by-step footprint. section 10.2.2
Figure 10.21 Vercel’s v0.dev discloses its CoT, that is, the model’s thinking process, so that the user can revert to problematic steps in a follow-up.
Vercel’s v0.dev illustrates a disclosed thinking process that users can revisit when a step appears problematic. section 10.2.2
