Google said its Search results will assemble custom layouts in real time, generating interactive visuals, tables, graphs, and simulations to fit a specific query, the company confirmed at its I/O conference on Tuesday. The capability matters because it lets Google build the answer’s presentation itself, which reduces the value of a publisher’s own page design as a reason for a searcher to click through.
The feature, which Google calls generative UI, is powered by Gemini 3.5 Flash and the agentic coding capability Google brings into Search from its Antigravity development environment. According to 9to5Google, which covered the keynote, Search can construct the response in the format the question calls for and render it on the fly. Google said the generative UI capability arrives for all Search users this summer at no cost, while the redesigned Search box that hosts it starts rolling out this week.
The mechanism is the important part for technical teams. To build an interactive comparison table or a working simulation, Google’s system needs structured, parseable data it can extract and recompose. A page that exposes its facts in clean semantic HTML, with accurate structured data and well-formed tables, is a page Google can read and reuse. A page that hides the same facts inside images, unlabeled JavaScript widgets, or decorative markup gives the system nothing to assemble.
That reframes a familiar trade-off. For years, a distinctive on-page experience, a custom interactive tool or a well-designed comparison, was a moat: it gave searchers a reason to visit and stay. Generative UI narrows that moat. If Google can rebuild the comparison inside the results page from a publisher’s underlying data, the searcher gets the interaction without the visit. The publisher’s design work still happened. The traffic from it may not.
Google has not detailed how it sources the data behind a generated layout, or whether the layout cites the pages it drew from. That is the unanswered question that decides the impact. A generative layout that links its sources is a new surface to optimize for. A layout that synthesizes from many pages without attribution is a new way for the work to be used without a return click. Google’s announcement covered the capability, not the citation model.
There is a defensive read and an opportunistic one. The defensive read: data a publisher considers proprietary, original research, pricing, structured comparisons, becomes easier for Google to extract and recompose once it sits in clean machine-readable markup. The opportunistic read: clean structured data is also what makes a page eligible to be the source Google draws from, and being the source is better than being invisible. The two readings are not contradictory. They are the cost and the entry ticket of the same change.
The build is the part Google did not discuss in depth: generated simulations and interactive widgets running inside results raise real questions about accuracy and freshness, since a recomposed answer is only as current as the data it pulled. Teams should not assume a generated layout reflects their latest numbers.
Technical SEO teams should audit how their highest-value data is exposed this quarter: facts and figures in semantic HTML and valid structured data are eligible to be sourced, while the same facts locked in images or opaque scripts are simply unreadable to Google’s new layout engine.
ATTRIBUTION: Reported by 9to5Google from the Google I/O 2026 keynote on May 19, 2026.
