If you opened ChatGPT this week and asked it to help plan a dinner party — and instead of a wall of text, you got a table you could actually edit — you're not imagining things. The product changed.

On October 7, OpenAI rolled out GPT-6 and a feature called Intelligent UI to ChatGPT. Paid subscribers (Plus, Pro, Business, Enterprise) got it first. Free and Go users followed October 8. What changed: ChatGPT now decides, question by question, whether to answer you in text or in something interactive — a clickable chart, a built-in calculator, an editable table, a tappable diagram.

Here's what that means in practice.

What actually changed

The easiest way to explain it is with a contrast.

Old behavior: you type "help me figure out how much food to buy for a dinner party," and ChatGPT writes back a paragraph. Maybe a bulleted list if it's feeling structured. You copy things, do the math yourself, try to adapt the answer to your situation.

New behavior: you type the same question and ChatGPT may respond with a form. You type in how many guests you're expecting, and the quantities update. No copying. No math on your end. The answer knows it's a variable.

OpenAI says the model decides for itself which format fits the question. Ask about a bicycle's parts, you might get an interactive diagram with clickable sections for the frame, wheels, and drivetrain. Ask for a recipe for six people, potentially get sliders to adjust the serving count. Ask a math question, potentially get a calculator that already understands the context you set up.

One thing worth understanding: ChatGPT isn't building you a custom app. The AI is generating these interactive elements on the fly, the same way it generates text. The output just looks and behaves differently than a paragraph. Think of it less like a software feature and more like the model learning a new vocabulary for how to express an answer.

Source spread

• aiweekly — OpenAI launches Intelligent UI alongside GPT-6 — builder. Covers the October 7 rollout, tier sequencing, and which models (Sol for paid, Luna for free).
• gHacks Tech News — OpenAI brings GPT-6 to all users — builder. Describes the interactive elements and the Settings toggle for turning them off.
• The Decoder — ChatGPT ditches text output for interactive UI — builder. Good overview with specific component examples, including the bicycle breakdown demo.
• Engadget — GPT-6 coming to all ChatGPT tiers — builder. Covers the rollout timeline and explains the model-name distinction by tier.

What's real

• This is the regular ChatGPT you use every day. Not the developer API. Not the code tools. The thing at chat.openai.com — or the app on your phone — that you ask questions to. That thing.
• The model still answers in text when a visual adds nothing. Ask a short factual question, you still get a short answer. The interactive format kicks in when the question is one where structure genuinely helps.
• There's an opt-out. If you prefer plain text, go to Settings on the web and turn off Layout and visuals. Some visual elements may still appear, but the toggle significantly reduces them.
• Paid and free users both get the feature. Paid subscribers run on GPT-6 Sol; free and Go users get GPT-6 Luna. Both support Intelligent UI, though Sol may generate more complex interactive elements in practice.
• Responses start faster, too. OpenAI claims web search queries begin responding an average of 44% faster than under GPT-5.6 Instant. One figure from their German site says 32%. Both numbers say the direction is right.

What deserves a side-eye

• The "model decides the format" framing sounds great until the model decides wrong. A reply that generates an interactive calculator you didn't want is a new kind of wrong answer. Not dangerous — just annoying.
• Some of this existed before. Interactive charts have been in ChatGPT since June 2026, and math/science interactive modules since March. The October rollout is a significant expansion, not the first version of the idea.
• You can't tell the AI "always give me text" short of the Settings toggle. To steer format on a per-question basis, say so in your prompt: "Give me this as plain text" works fine.
• The AI-generated calculators and forms are still AI-generated. A calculator widget that ChatGPT built can contain math errors. For anything financial or medical, verify the underlying numbers independently.

What to do about it

Concrete steps, none requiring technical knowledge:

• Try it on a question with variables. Trip planning, party hosting, comparing two options side by side, anything that a table or a calculator would make clearer. See if the format actually helps or gets in your way. Your own answer to that is the only review that matters for you.
• If you hate the visual mode, turn it off. Settings › Layout and visuals. Under your profile menu on the web. Won't fully remove visuals, but significantly reduces them.
• If the format is wrong, just say so. "Give me this as a plain text list." "Can you put this in a table instead?" The model follows formatting instructions well.
• Cross-check anything the calculator produces. The interactive elements are generated by an AI that can make arithmetic errors. For anything financial, medical, or otherwise consequential — verify the underlying math yourself.

Further reading

• OpenAI — GPT-6 announcement — primary company framing of the model and the Intelligent UI feature
• gHacks Tech News — OpenAI brings GPT-6 to all ChatGPT users — concise feature overview with opt-out instructions
• The Decoder — ChatGPT ditches text output for interactive UI — best walkthrough of the specific interactive element types
• RuntimeWire — OpenAI launches Intelligent UI — context on the feature's development history and earlier interactive releases