Generative UI in plain English.

Generative UI is a model deciding not only what to say but what interface to show. Instead of describing a comparison in prose, it renders a sortable table; instead of listing options, it offers a form; instead of explaining a calculation, it builds a small calculator you can adjust. OpenAI brought this to every ChatGPT user on October 7 and 8, 2026 as "Intelligent UI" with the GPT-6 rollout, with a setting to reduce interactive responses for people who prefer plain text. The idea builds on earlier pieces: tool calling (the model invokes a component), structured output (it emits data a renderer can draw), and artefact panels in Claude and others. The benefit is answers you can manipulate rather than read; the costs are consistency (every reply can look different), accessibility (generated widgets must still work with screen readers and keyboards), and trust (a confident-looking chart is still only as good as the model behind it). For product teams it changes what "the UI" means: much of it is now produced at runtime.