On October 7, 2026, OpenAI began rolling GPT-6 out to ChatGPT users with a feature it calls Intelligent UI. The idea behind Intelligent UI is that ChatGPT stops answering mostly in paragraphs and instead returns interactive interfaces: tappable buttons, editable charts, task-specific calculators and small apps built inside the conversation. It is the clearest sign yet that generative UI, where a model designs the screen as well as the answer, is becoming the default shape of a chatbot rather than an experiment.
OpenAIWhat OpenAI shipped: Intelligent UI explained
According to TechCrunch, Intelligent UI adds visuals to conversations, many of them interactive: tappable buttons, calculators built for the task at hand, and charts and graphs that users can edit. OpenAI's stated aim is making complex topics easier to learn. Product manager Aarush Selvan told TechCrunch that "ChatGPT has predominantly been a text-based interface." Demo examples included a diagrammed explanation of how an airplane wing generates lift, recipe visuals, a bicycle mechanics diagram, a map for a multi-day hiking trip and a personal savings calculator.
The Decoder adds that the format adapts to the question: a comparison can appear side by side, and an explanation can become an interactive chart. Users can also build small tools in the chat, such as a savings calculator or a game. TechCrunch notes that users can reduce the number of visuals in responses, so the text-first experience is not removed.
One more change is separate from the interface. The Decoder reports that GPT-6 can respond while it is still reasoning, and that OpenAI says this cuts wait times by 44 percent and can improve answers. The Decoder also reports that, in OpenAI's internal tests, GPT-6 beat its predecessor GPT-5.6 on difficult web searches using this method. These are OpenAI's own figures, reported by the press, and we have not seen independent measurements.
Rollout by tier: who gets which GPT-6 model
The rollout is global and staggered by a day. Paid plans get GPT-6 Sol and the free tier gets GPT-6 Luna, per The Decoder.
GPT-6 and Intelligent UI: who got it, and when
Global rollout dates by ChatGPT plan, with the GPT-6 model each group receives.
Two details are worth flagging. First, neither article we could read states which model the lower-cost Go plan receives, so we have left that open. Second, a Neowin headline describes the launch as reaching 12 billion ChatGPT users. We could not open that article, and the figure does not appear in the TechCrunch or Decoder coverage, so we are not relying on it. For model details, see our earlier coverage of GPT-6 Sol and Luna and the GPT-6.1 Sol release.
Generative UI as a design shift
Chat interfaces began as a transcript: you type, a model types back. Several products have since moved the output into a richer surface, and Intelligent UI fits a clear lineage.
- June 2024, Claude Artifacts. Anthropic introduced Artifacts alongside Claude 3.5 Sonnet on June 21, 2024. Content such as code or documents appeared in a dedicated window next to the conversation, where users could review and edit it.
- November 2025, Gemini dynamic view. Google Research described generative UI on November 18, 2025. In the Gemini app's dynamic view, Gemini designs and codes a customized interactive response for each prompt. Google also put it in AI Mode in Search for AI Pro and Ultra subscribers in the US.
- May 2026, Gemini Neural Expressive. The Decoder says Google added a similar feature to Gemini in May under the name Neural Expressive. Secondary sources describe it as a broader redesign of the Gemini app announced around Google I/O 2026, so the exact scope is less clear than the earlier items.
- October 2026, ChatGPT Intelligent UI. The difference is distribution. This reaches every ChatGPT plan, including the free tier, within two days.
The mechanism matters for understanding the shift. In the artifact model, the person asks for a thing (a document, an app) and the model produces it in a side panel. In the generative UI model, the model decides on its own that a question deserves a chart or a form and renders one inline. The interface becomes an output of the model, not a fixed container around it. That moves a design decision, which control suits this answer, from a product team to a model that makes the call per response.
In generative UI, the interface stops being a fixed container and becomes part of the answer.
Metir AI analysis
Implications for publishers and apps
None of the sources we reviewed describes how Intelligent UI treats third-party content or links, so what follows is analysis, not reporting.
- Answers that are already interactive. If a calculator, comparison table or map is generated inside the chat, some questions that once sent a reader to a specialist site may now be resolved in place. That is a pattern to monitor in referral traffic, not a measured effect.
- Mini apps as a lightweight app layer. The Decoder's report that users can build small tools in the chat points to disposable, single-purpose software. It overlaps with what spreadsheets, form builders and simple utilities do today.
- Structured data may matter more. A model that renders a chart needs clean numbers to draw it. Sources that publish well-structured, citable data are easier to turn into visuals than long prose.
- Developers. OpenAI's announcement and system card are cited by The Decoder, but we could not open OpenAI's own page, so whether Intelligent UI is exposed through the API is unconfirmed.
Accessibility and control
Interactive output raises accessibility questions that coverage so far leaves unanswered. TechCrunch suggests the format could make ChatGPT more accessible and reach a wider audience, though it says OpenAI did not state this directly. That is plausible: a diagram can help a visual learner, and a form can be easier than a long instruction. The risks run the other way too. Charts and custom controls need text alternatives, keyboard navigation and screen reader labels, and a model generating those on the fly makes consistency harder to guarantee than a fixed interface does. The option to reduce visuals is a useful control, but whether it covers assistive-technology needs is not documented in the reports we found.
What to watch next
- Independent latency tests. The 44 percent wait reduction is OpenAI's claim. Third-party timing across tiers would show how it holds up in practice.
- Free-tier behavior. Luna users get the interface on October 8, and how well a smaller model chooses and builds visuals is an open question.
- Accessibility documentation. Whether OpenAI publishes how generated interfaces handle screen readers and keyboard use.
- Publisher and developer terms. Any guidance on attribution, linking or API access for the generated interfaces.
- Competitive response. Google's Gemini already ships generative UI, and Anthropic's Artifacts predate both, so convergence on similar patterns is likely.
For readers who switch between models, the practical point is that interface styles now differ by provider as much as answer quality does. Metir gives access to several providers in one place, so you can compare how each handles the same question.
Sources:
- TechCrunch: ChatGPT is getting a lot more visual, with the launch of a new interface (Oct 7, 2026)
- The Decoder: ChatGPT with GPT-6 ditches mostly text output for interactive UI (Oct 7, 2026)
- Neowin: OpenAI brings GPT-6 to ChatGPT users with new Intelligent UI (headline only, page not accessible to us)
- Anthropic: Claude 3.5 Sonnet and Artifacts (June 2024)
- Google Research: Generative UI (Nov 18, 2025)
Image credits
- OpenAI headquarters at 1515 Third Street, San Francisco: photo by Coolcaesar, licensed CC BY 4.0, via Wikimedia Commons.