OpenAI is changing the basic shape of a ChatGPT answer. Instead of responding mainly with paragraphs and occasional images, the company’s new “Intelligent UI” can generate custom visual and interactive elements when the system determines that they would make an answer more useful, according to an October 7 report from WIRED.

The experience is powered by GPT-6. WIRED reported that it began rolling out to paid users on October 7, with free users scheduled to receive it the following day. OpenAI is not promising a graphic for every prompt; the goal is to select diagrams, controls, and other interactive components only when they add more value than another block of prose.

Three abstract interactive stations represent anatomy, housing affordability, and airplane seat exploration.
WIRED tested generated experiences ranging from an anatomy diagram to an affordability calculator and a seat explorer.

WIRED’s hands-on examples show how broad that idea can become. A request about slug anatomy produced a visual with buttons for different body parts. A San Francisco apartment-affordability question became a calculator with sliders for income and expenses. A comparison of four Alaska and Delta aircraft layouts generated a seat explorer that let the reviewer select individual seats and identify exit rows.

These outputs are more than decorated answers. Each one behaves like a small, purpose-built software tool assembled around a specific question. The affordability calculator invites the user to change assumptions, while the aircraft explorer supports comparison and inspection. That makes the conversation less like reading a report and more like manipulating a model of the problem.

OpenAI product manager Aarush Selvan told WIRED that the company’s design team worked on deciding when a diagram, chart, or set of buttons would be helpful and when those additions would create clutter. Users who prefer simpler responses can also ask ChatGPT to generate fewer visual elements, making the interface itself partly responsive to personal preference.

One question branches into multiple personalized interface layouts made from rearranging controls.
Generative interfaces may personalize software for each question while challenging the idea of one fixed design.

OpenAI is not alone in this direction. Google has described comparable custom Search results as “generative UI.” WIRED cited an adjustable black-hole explainer as one example of Google’s approach. Both systems point toward an interface that is produced in response to the user rather than selected from a fixed collection of screens designed in advance.

The shift also complicates the role of software designers. If an AI system decides what controls, layouts, and visual explanations to generate for each person, the experience may no longer have one stable interface. WIRED noted that this could create an identity challenge for designers as AI takes a larger role in determining what software looks like.

For ChatGPT’s audience of more than a billion users, Intelligent UI could make many answers easier to explore, but the quality of those generated tools will matter as much as their novelty. The feature’s real test will be whether its temporary interfaces clarify decisions without adding confusion—and whether users can trust the calculations, labels, and comparisons those interfaces present.