Generative UI Is Live in Google and ChatGPT — and No Website Is Inside It

By Tom Whitfield

Google's generative UI reached AI Mode globally in English on Aug. 19, 2026, and OpenAI launched its own version, Intelligent UI, in ChatGPT on Oct. 7, 2026. Neither company has shown a single website's data inside a generated interface. That gap is the story for marketers, and it mirrors the decision hotels faced when Booking.com became the channel that owned the customer.

What has actually shipped?

Generative UI means the AI builds the screen for each answer — charts, buttons, maps, small tools — instead of returning a block of text or a list of links.

The Google timeline is short and dense. Google first brought generative UI to AI Mode with Gemini 3 on Nov. 18, 2025. At I/O on May 19, 2026, it promised the capability to everyone. Its Search post from I/O says Search "can design custom layouts, assembling components (like interactive visuals, tables, graphs or simulations)," and that these capabilities "will be available for everyone in Search this summer, free of charge."

The Aug. 19 post confirmed the rollout: "Generative UI capabilities have launched globally in English in AI Mode" and "have started rolling out as well in AI Overviews," the AI summaries at the top of Google results. The examples in that post are study aids — "custom tools and simulations to help you see complex topics in action."

Google also explained how the screens are built. Its I/O post says: "We're bringing the power of Google Antigravity and the agentic coding capabilities of Gemini 3.5 Flash right into Search." Antigravity is Google's tool for AI that writes code. Google's November 2025 post says Gemini codes the interactive tools in real time and adds them into the response.

OpenAI's approach differs in construction. ChatGPT answers can now "include graphics, tappable buttons, forms, charts, and interactive experiences you can use directly in your conversation," the company said. OpenAI rolled Intelligent UI, launched with GPT-6, to most paid ChatGPT plans on Oct. 7 and to the Free and Go plans the next day. ChatGPT assembles its screens from ready-made parts while the answer is still being written; Google says Gemini writes the code on the spot.

OpenAI's examples: a Sunday roast timed alongside the recipe, a road trip on a map, the central limit theorem, a bill splitter, a retirement calculator. Its announcement closes with a line that states the stakes plainly: "Instead of people adapting to software, software will adapt to people." The Intelligent UI post mentions web search but says nothing about websites, sources, or links back.

What's missing from every demo?

Every example in both companies' announcements is built from what the AI already knows. None shows a store's product, a hotel's rooms, or a local business's hours inside a generated interface.

Google's one I/O example that uses outside information is a later feature: small apps people will be able to build inside Search "in the coming months." The example is a fitness tracker. "Search will code it for you, tapping into fresh, real-time sources including reviews, live maps and local data like the weather," the post says. It does not say whose reviews.

Google's November 2025 post says "all responses have prominent links to high-quality content across the web." That puts links next to the generated interface — not a business's own data inside one. Observers have seen rich results with extras like star ratings or prices, but there is no way to confirm those were generative UI.

How many people see this?

Usage of the underlying products is enormous; exposure to generated interfaces is unquantified.

• AI Overviews: "over 2.5 billion monthly active users" (Google post of June 3, updated Aug. 31)
• AI Mode: past 1 billion monthly users (same post)
• ChatGPT: more than 1.2 billion weekly users (OpenAI)

Neither company has said how many people actually see a generated interface. One ChatGPT Plus user reported on OpenAI's forum that their account showed GPT-6 but no Intelligent UI. That is a single report.

Why the platforms want you to stay

Generative UI keeps the user with Google longer, or all the way through to a transaction. And it is not just Google — it is every platform that builds it.

There is a second driver, and the author of the original analysis flags it as his own inference. On Alphabet's earnings call on July 22, Sundar Pichai said Google has started training Gemini 4. Asked by an analyst how Google would close the gap in coding, Pichai answered: "There are areas where we have acknowledged we need to improve. Coding and agentic coding is an example of that, and I think the teams are very, very focused on it." On Sept. 30, Pichai showed an early version, Gemini 4 Argon, on X and said it performs at the top level in "software engineering," based on Google's own tests. Since Gemini codes the tools inside generative UI, a Gemini that is better at coding builds better tools.

Will Google become your Booking.com?

A business facing generative UI must decide what to do about Google traffic: wait for the booking or the order to arrive via Google — a version of what Booking.com does — or build other channels. Or do both.

What a business would pay Google for those bookings is unknown. The original analysis's hunch: there is zero chance Google is not thinking about how to monetize generative UI, likely through a commission, paid placements, or both — essentially the Booking.com business model on a search results page.

The paid-placement half already exists. At Google Marketing Live in 2025, Google said, "We're expanding Ads in AI Overviews, and testing Ads in AI Mode." Paid placement inside Google's AI answers was already a product by 2025. A commission on bookings made inside a generated interface is not something Google has announced.

Nothing Google or OpenAI has shown puts a website inside a generated interface yet. The choice between waiting for Google to send the booking and building the channels a business owns does not need to wait for that.