ChatGPT has answered every question the same way since it launched in 2022: a block of text. On October 8, 2026, that stopped being universally true. OpenAI extended a feature called Intelligent UI to every ChatGPT tier, including Free and Go — meaning the chatbot can now respond to a question with tappable buttons, adjustable forms, live charts, or a small working tool, instead of just a paragraph describing one.
The rollout itself started a day earlier, on October 7, for paid Plus, Pro, and Business accounts, with Enterprise timing set by each workplace's admins. Free and Go users — by far the larger slice of ChatGPT's reported 1.2 billion weekly users — got it October 8, which is the more significant date if you're measuring by how many people the change actually reaches.
Quick summary: OpenAI's Intelligent UI began reaching every ChatGPT tier on October 7-8, 2026 — Plus, Pro, and Business (running GPT-6 Sol) got it October 7, with Enterprise timing set by workplace admins; Free and Go (running GPT-6 Luna) got it October 8. Instead of pure text, ChatGPT can now generate interactive elements — charts, diagrams, tappable buttons, adjustable forms, checklists, maps, and small built-in tools like bill splitters or calculators — built from what OpenAI describes as a library of native, streamable components compiled as the model generates its answer. OpenAI also says GPT-6-powered web search now answers 44% faster on average than GPT-5.6 Instant. The feature is limited to ChatGPT's Chat tab (Work and Codex are unaffected) and can be turned off in web settings. Reaction has been split, with the skepticism centered on trust, consistency, and security risks that come with letting a model generate interactive, clickable interfaces rather than inert text.
What Intelligent UI actually does
OpenAI's own demo examples give a clearer sense of this than the name does. Ask about how a bicycle's gears, brakes, and drivetrain work together, and instead of five paragraphs, you get an interactive breakdown of the five connected systems you can click through individually. Ask for help planning a Sunday roast for six guests, and you get dish images, a shopping list calculator that adjusts quantities if a guest count changes, and a cooking timeline checklist — not a wall of prose you'd have to re-read while standing in the kitchen. OpenAI says people are already using it to spin up small working tools inside a conversation: savings calculators, bill splitters, even simple playable games (Sources: MacRumors, Unite.AI).
Mechanically, OpenAI describes the feature as running on "a library of native, streamable components with a compiler that processes the interface as the model generates it" — meaning the interface builds progressively while the model is still reasoning, rather than waiting for a complete response before rendering anything. That's also why OpenAI can claim a responsiveness improvement alongside a visual one: the system starts streaming an answer sooner and lets you interact with parts of it before the full response finishes generating (Source: redreamality.com).
Who gets it, and when
| ChatGPT Tier | Model Underneath | Intelligent UI Reached This Tier |
|---|---|---|
| Plus / Pro / Business | GPT-6 Sol | October 7, 2026 |
| Free / Go | GPT-6 Luna | October 8, 2026 |
| Enterprise | GPT-6 Sol | Timing depends on each workplace's admin settings |
A few scope limits are worth knowing before you go looking for this everywhere. Intelligent UI only shows up in ChatGPT's regular Chat tab — the separate Work and Codex surfaces are unaffected. It's rolling out gradually across the web app and updated mobile apps, and OpenAI has said older macOS and Windows desktop app versions miss out for now, so an outdated desktop install may not show the feature even on a paid plan. You can also turn it off entirely in web settings if you'd rather keep getting plain text. GPT-6 itself replaces the "Latest" option in ChatGPT's model picker, though OpenAI left GPT-5.6 Sol available to switch back to for comparison (Sources: Android Headlines, MacRumors).
The part that's dividing people
Early reaction to the rollout, including a long Hacker News discussion, has been genuinely split, not a clear win. The skepticism isn't really about whether interactive answers are useful (most people seem to agree they can be); it's about trust. A usable-looking button, form, or calculator can make an answer feel more authoritative than the same information in a paragraph would, even when the underlying content is wrong — and unlike a factual error in a sentence, a broken or misleading interactive element is harder for a reader to simply disbelieve on sight (Sources: Hacker News, explainx.ai).
There's a sharper security version of that concern, too, sometimes called the "lethal trifecta" — a term coined by independent researcher Simon Willison for the combination of private data, untrusted content, and a channel for external communication, which together let an AI system be tricked into leaking information it shouldn't. A generated interface that can post messages, submit forms, or embed images and links is exactly the kind of new "external communication" channel that concern is about: in principle, a cleverly crafted interactive element could smuggle conversation data out to an attacker-controlled destination, or present a phishing-style button disguised as a legitimate part of the interface. Separately, at least one study cited in coverage of the rollout found that 55.8% of LLM-generated e-commerce-style interface components contained deceptive design patterns, a figure that rose by another 15.8 percentage points when the model was prompted under simulated sales pressure — evidence the risk isn't purely theoretical (Source: redreamality.com).
Independent analysis of OpenAI's own system card for the release has also pointed out a gap: there's reportedly no safety evaluation specific to generated interfaces included alongside the launch, as distinct from the text-output safety testing OpenAI already does. That doesn't mean Intelligent UI is unsafe — it means the newest category of risk it introduces hasn't yet gotten its own dedicated evaluation, at least not one OpenAI has published (Source: redreamality.com).
Until Intelligent UI has more of a track record, treat anything that asks you to type in sensitive information — passwords, card numbers, personal details — inside a ChatGPT-generated form or button with the same skepticism you'd apply to an unexpected pop-up on a website. A legitimate-looking interactive element generated mid-conversation is not the same as a form on a site you already trust.
- Everyday ChatGPT users on any tier, including Free — this changes what responses look like by default, not just for paying subscribers.
- Anyone building on top of ChatGPT or evaluating it for a business workflow where interface trust and auditability matter.
- Security-conscious users who want to know there's a real, named risk category (the 'lethal trifecta') behind the 'be careful what you click' advice, not just general caution.
- Less urgent for Work and Codex users — this specific rollout doesn't touch those surfaces yet.
01When did ChatGPT's Intelligent UI roll out to free users?
October 8, 2026. Plus, Pro, and Business got it a day earlier, on October 7; Enterprise timing depends on each workplace's admin settings.
02What is Intelligent UI?
A GPT-6 feature that lets ChatGPT answer with interactive elements — charts, diagrams, tappable buttons, adjustable forms, checklists, maps, and small built-in tools like calculators or bill splitters — instead of only text, built from a library of components the model assembles as it generates its answer.
03Which model powers Intelligent UI on each ChatGPT tier?
GPT-6 Sol powers Plus, Pro, Business, and Enterprise. GPT-6 Luna powers the Free and Go tiers.
04Can I turn off Intelligent UI?
Yes — OpenAI says it can be disabled in ChatGPT's web settings if you'd rather keep getting plain-text responses.
05Is Intelligent UI safe to use for things like payments or personal data?
Treat it cautiously. Independent analysis has flagged that generated interfaces create a new risk category — sometimes called the 'lethal trifecta' of private data, untrusted content, and external communication — and that OpenAI's system card reportedly doesn't yet include a safety evaluation specific to generated interfaces. Avoid entering sensitive information into a ChatGPT-generated form or button the same way you'd be cautious with an unexpected website pop-up.
Whether you find Intelligent UI genuinely useful or mostly a reason for extra caution probably depends on what you use ChatGPT for — and the only real way to find out is to see how it behaves on your own questions. LumiChats gives you access to 40+ premium AI models, including the latest from OpenAI, Anthropic, Google, and xAI, side by side in one place, so you can compare how different models actually respond to the same prompt before deciding which interface — or which model — you trust with it.
