Generative UI Notes

What Is Generative UI, and Why Is It Controversial?

Generative UI (GenUI) is gaining traction as a possible future direction for web design. Products like Figma Sites already hint at it, letting users describe what they want and getting a site built by AI. The concept turns conventional interface design inside out: instead of anticipating what users might need and building interfaces to accommodate those needs, generative AI observes the user and generates a custom interface in real time. In this model, every visit to a site can render a different experience — a bespoke interface unique to that moment.

This is not a settled technology. It is a developing field with open questions about reliability, accessibility, and what it means for designers. What follows is a working set of notes on the current state of GenUI, its definitions, its tools, and the controversies around it.

Defining GenUI

Most definitions of GenUI center on real-time generation of complete interfaces, not just content. Google Research's team describes it this way:

Generative UI is a new modality where the AI model generates not only content, but the entire user experience. This results in custom interactive experiences, including rich formatting, images, maps, audio and even simulations and games, in response to any prompt (instead of the widely adopted "walls-of-text").

The Nielsen Norman Group provides a more user-focused definition:

A generative UI (genUI) is a user interface that is dynamically generated in real time by artificial intelligence to provide an experience customized to fit the user's needs and context.

UX Collective's framing is broader, emphasizing inputs and context as the source of adaptation:

A Generative User Interface (GenUI) is an interface that adapts to, or processes, context such as inputs, instructions, behaviors, and preferences by using generative AI models (e.g. LLMs) in order to enhance the user experience.

Generative Versus Predictive AI

The term "AI" gets used casually to cover very different technologies. A useful distinction separates generative AI from predictive AI:

Generative AI does not just forecast what is likely next; it generates new output from learned patterns. Within that capacity, GenUI is the specific application of generating an interface based on what the model knows about the user in real time.

Accessibility: The Core Criticism

The bigger unresolved issue is whether all of this benefits all users. The current crop of so-called AI website builders is technically different from true adaptive GenUI — those builders do not customize for each visitor. But the trajectory points that way, and the accessibility questions are already urgent.

What happens when an interface is generated at runtime? Is there a way to generate it to accommodate users with aural, visual, or physical impairments? Early results are not encouraging. A high-profile example is Figma Sites, which shipped a GenUI web builder that drew across-the-board criticism for accessibility failures, most notably from Adrian Roselli. Figma has since pushed updates designed to address these issues and published a dedicated guide to improving accessibility on generated sites, but the fixes stop short of full remediation.

Figma is not the only company heading this way, though. WordPress has an AI site builder, and Vercel, Squarespace, Wix, GoDaddy, Lovable, and Reeady all offer their own takes on AI-generated web development.

The gap between working tools and an accessible experience has also produced dispute about whether GenUI could replace accessibility experts. Jakob Nielsen claimed in 2024 that generative UI would solve accessibility by providing individualized UX far beyond what human practitioners can do. The claim met substantial pushback from accessibility and UX specialists, and while Nielsen later walked back parts of it in an updated piece, the position remains contested.

This accessibility gap is less about a moral stance and more about evidence, or its absence. Even Google's own People + AI Guidebook, built on extensive usability research, does not mention accessibility in its best practices despite constant reference to human-centered design. If GenUI is going to be a genuine future for the web, accessibility has to be treated as a core component of generation, not an afterthought layered on later.

Real-World Tools and Projects

Despite the critique, development is moving quickly, and several notable projects exist.

Google maintains an open site dedicated to GenUI research, offering concrete examples of how a prompt can produce diverse user interfaces. The work continues further with Project Genie, an invite-only prototype focused on generating so-called "interactive worlds" on the fly.

For developers, Google also provides a GenUI SDK that plugs into Flutter applications. It is an explicit tool to connect your app to an LLM provider and pass dynamic UI generation capability to your app's interface layer.

Adaptive, runtime-generated interfaces are also an area of focus for a few startups. Thesys makes generative experiences a control layer for SaaS products, and Copilot provides a generative UI playground as a developer technology for building component trees on the fly.

Perhaps the most important caveat is that not all GenUI is made equal. A website generator that produces a static, one-size-fits-all page after a single prompt is a very different thing from a system that renders different components, layouts, and styles based on every visit. Both are called "generative," but only the second truly fits the definition of user-adaptive GenUI. As the tools mature, that distinction matters a great deal when evaluating how a site is likely to treat accessibility and user difference.