ChatGPT is learning to answer with more than words. OpenAI has launched “Intelligent UI”, a feature that lets the chatbot build interactive elements into its replies: tappable buttons, forms, charts, calculators and diagrams. It arrives with GPT-6 and began rolling out globally on 7 October.
What it does
OpenAI says its GPT-6 models will produce “graphics, tappable buttons, forms, charts, and interactive experiences” when they suit the answer. In the demonstrations, a question about how a bicycle works produced a diagram of a seven-speed bike with a button for each part, which you can tap to read more. A plan for a roast lamb dinner came with a menu of dishes and a button to change the number of guests, which rescaled the ingredients. TechCrunch lists other examples, including a diagram of how an aeroplane wing generates lift, recipe visuals, a map of a multi-day hike and a savings calculator, and says the charts can be edited.
OpenAI says the aim is to make complex subjects easier to learn. Its announcement lists four uses: exploring ideas with interactive visuals, creating one-off tools such as a game, a calculator or a bill splitter, learning concepts by changing an input to see the effect, and adjusting an answer with controls built into the response. A comparison might appear side by side and an explanation might become an interactive diagram, while a plain text answer is still given when that is most useful.
How it decides
According to Gizmodo, ChatGPT creates these interfaces by itself when it judges them the best way to answer, drawing on a library of components the model was trained to use. Simple questions still get a text answer. You can also ask for an interface whenever you want one, and you can turn down the amount of visuals, in the way you can adjust ChatGPT’s personality settings.
Who gets it, and when
The rollout reached ChatGPT Pro, Plus, Business and Enterprise users on Wednesday 7 October, and free and Go users the next day. According to OpenAI’s announcement, paid plans get GPT-6 Sol and free and Go users get GPT-6 Luna. OpenAI also says the interface renders progressively as the model generates it, and that GPT-6 Instant starts answering web-search questions 44% sooner on average than GPT-5.6 Instant. That measures time to the first answer, not the time to finish. On Enterprise plans, workspace administrators have to switch it on. Neither report mentions any difference by country.
What to watch
Both reports rely on OpenAI’s own demonstrations, so how well it works day to day is unproven. Gizmodo’s reporter flags two things to watch: that interactive answers could use up your usage allowance faster, and that visual elements add a new place for the chatbot to get things wrong, since a confident-looking chart can still contain an error. He also notes that Google introduced a similar “Generative Interface” with Gemini 3, which created buzz but did not clearly set Google apart.
If you try Intelligent UI, check any numbers in a chart or calculator against a reliable source before you rely on them. We have not tested the feature ourselves and will update this article after independent testing.
Leave a Reply