AI UX Patterns
Common interface patterns — streaming text, tool-call indicators, citations, and regenerate controls — that make AI output feel responsive and trustworthy.
Overview
Because model output is probabilistic and can be slow or wrong, AI interfaces have converged on a small set of recurring patterns that manage a user’s expectations — showing progress, showing sources, and making it easy to try again.
Where It Fits
Loading Indicator
Streaming Text
Tool-Call Indicator
e.g. "Searching…"
Citations
Regenerate Control
Key Points
- Progressive disclosure
- Showing "Searching…" or "Reading document…" during a tool call keeps a slow multi-step response feeling responsive.
- Citations
- Surfacing the sources behind a RAG-generated answer lets a user verify it rather than trust it blindly.
- Regenerate / try again
- Because output is probabilistic, an easy way to re-run a request is a near-universal AI UI pattern.
Interview Question
Why do most AI chat interfaces show a “searching” or “thinking” indicator instead of a plain loading spinner?
A multi-step response — retrieving documents, calling a tool, then generating — can take several seconds. A generic spinner gives no information, while a step-specific indicator sets accurate expectations and reduces the perceived wait, which matters more for AI responses since latency is inherently variable.
Explain It in 30 Seconds
AI interfaces converge on a small set of UX patterns — streaming text, step-specific loading indicators, citations for generated claims, and an easy regenerate control — because model output is slower and less predictable than a typical API response.