I’m actually on this rising concept that the way forward for internet design is Generative UI Design. We see hints of this already in merchandise, like Figma Websites, that tout having the ability to create web sites on the fly with prompts.
Placing apart the clear downsides of transport half-baked know-how as a production-ready product (which is difficult to do), the angle I’m notably is analysis geared toward utilizing Generative AI (or GenAI) to output customized interfaces. It’s wild as a result of it utterly flips the best way we take into consideration UI design on its head. Slightly than anticipating consumer wants and designing round them, GenAI sees the consumer wants and produces an interface custom-tailored to them. In a way, a web site turns into a snowflake the place no two experiences with it are the identical.
Once more, it’s wild. I’m not right here to take a position, opine, or preach on Generative UI Design (let’s name it GenUI for now). Simply free notes that I’ll replace as I proceed studying about it.
Defining GenUI
Google Analysis (PDF):
Generative UI is a brand new modality the place the AI mannequin generates not solely content material, however all the consumer expertise. This leads to {custom} interactive experiences, together with wealthy formatting, pictures, maps, audio and even simulations and video games, in response to any immediate (as a substitute of the broadly adopted “walls-of-text”).
A generative UI (genUI) is a consumer interface that’s dynamically generated in actual time by synthetic intelligence to supply an expertise custom-made to suit the consumer’s wants and context.
A Generative Person Interface (GenUI) is an interface that adapts to, or processes, context resembling inputs, directions, behaviors, and preferences via the usage of generative AI fashions (e.g. LLMs) to be able to improve the consumer expertise.
Put merely, a GenUI interface shows totally different parts, data, layouts, or types, primarily based on who’s utilizing it and what they want at that second.
Generative vs. Predictive AI
It’s simple to dump “AI” into one large bucket, but it surely’s usually distinguished as two differing types: predictive and generative.
| Predictive AI | Generative AI | |
|---|---|---|
| Inputs | Makes use of smaller, extra focused datasets as enter information. (Smashing Journal) | Educated on giant datasets containing thousands and thousands of pattern content material. (U.S. Congress, PDF) |
| Outputs | Forecasts future occasions and outcomes. (IBM) | New content material, together with audio, code, pictures, textual content, simulations, and movies. (McKinsey) |
| Examples | ChatGPT, Claude | Sora, Suno, Cursor |
So, after we’re speaking about GenAI, we’re speaking concerning the capacity to create new supplies educated on current supplies. And after we’re speaking particularly about GenUI, it’s about producing a consumer interface primarily based on what the AI is aware of concerning the consumer.
Accessibility
And I ought to word that what I’m speaking about right here just isn’t strictly GenUI in how we’ve outlined it as far as UI output that adapts to particular person consumer experiences, however fairly “growing” generated interfaces. These so-called AI web site builders don’t adapt to the person consumer, but it surely’s simple to see it heading in that route.
The factor I’m most curious about — involved with, frankly — is to what extent GenUI can reliably output experiences that cater to all customers, no matter impairment, be it aural, visible, bodily, and so on. There are loads of totally different inputs to contemplate right here, and we’ve seen simply how terrible the early outcomes have been.
That final hyperlink is an enormous poke at Figma Websites. They’re simple to poke as a result of they made the biggest business push into GenUI-based internet improvement. To their credit score (maybe?), they obtained the extreme pushback and determined to do one thing about it, asserting updates and publishing a information for enhancing accessibility on Figma-generated websites. However even these have their limitations that make an effort and recommendation appear much less helpful and extra about saving face.
Anyway. There are many different gamers to leap into the sport, notably WordPress, but in addition others like Vercel, Squarespace, Wix, GoDaddy, Lovable, and Reeady.
Some of us are extra optimistic than others that GenUI just isn’t solely able to producing accessible experiences, however will change accessibility practitioners altogether because the know-how evolves. Jakob Nielsen famously made that declare in 2024 which drew fierce criticism from the group. Nielsen walked that again a yr later, however not a lot.
I’m not even remotely certified to supply greatest practices, opine on the way forward for accessibility apply, or speculate on future developments and capabilities. However as I take a look at Google’s Individuals + AI Guidebook, I see no point out in any respect of accessibility regardless of dripping with “human-centered” design rules.
Accessibility is a lagging consideration to the hype, no less than to me. That has to vary if GenUI is actually the “future” of internet design and improvement.
Examples & Sources
Google has a repository of examples displaying how consumer enter can be utilized to render a wide range of interfaces. Going a step additional is Google’s Undertaking Genie that payments itself as creating “interactive worlds” which can be “generated in real-time.” I couldn’t get an invitation to strive it out, however possibly you may.
Along with that, Google has a GenUI SDK designed to combine into Flutter apps. So, yeah. Hook up with your LLM supplier and let it rip to create adaptive interfaces.
Thesys is one other one within the adaptive GenUI house. Copilot, too.
