UI for the masses
In the world of AI with everything, some applications make more sense (coding, personalisation, content for marketing) than others (I have yet to see one of those LinkedIn automations actually writing a post that is anything other than sheer garbage), the prospect of help when designing user interfaces for web and mobile platforms seems to be a no-brainer for the first category. Stitch, announced at Google I/O 25 is available in Beta and shows a great deal of promise, but there are some limitations that mean it will not be the answer to all of a busy designer's daily issues.
As with most of the recent tools released by Google that emply Gemini (Veo, Whisk, Gemini itself), the interface is incredibly simple, relying on Gemini to do most of the heavy lifting for you. When you log onto the site, currently only available in beta, you are faced with a simple input with a couple of options where you describe the kind of screen you want in plain language: from this, Stitch pretty much instantly generates a ready-to-use design. You can also upload a sketch or screenshot, and the app will turn it into a working interface. For someone who has no real experience of user interface (UI) design, this will look very impressive: if you have done much work previously, in particular combining UI with user experience (UX) work, it will range from pretty decent to... okay.
Stitch in Flash mode can produce user interfaces in seconds, but the real work lies in getting the right prompt.
How does it work?
Stitch works in a very natural and simple way. You can type a short description such as: “Make me a login page with a dark background and rounded buttons,” and in seconds you’ll get a clean design. If you have a rough drawing or an existing screenshot, you can upload it - although that onlly works in experiemental mode - and Stitch will create a polished version you can edit. As with Gemini, Google provides users with two choices for the AI models inside Stitch. The Flash model is fast and great for quick testing, while the Pro model produces more detailed results but is slower and has stricter limits on how many screens you can generate each month.
Once you have a design, Stitch lets you see different versions so you can pick the one you like best. You can also refine the results step by step. For example, you might ask it to “change the header to blue” or “add a search bar,” and it will update the screen instantly. When you’re happy, you can export the design into Figma for collaboration, or export it as HTML with CSS to use in development.
Easy to use - but with annoying limitations
The biggest advantage of Google Stitch is how much faster it makes prototyping. Instead of spending hours making wireframes or mockups, you can create a usable design in minutes. It also bridges the gap between designers and developers, since the exported code is clean enough to be used as a starting point in real projects. Another strong point is accessibility: even people with no design or coding background can create professional-looking screens, which makes it great for startups, students, or small teams who want to move quickly without hiring a full design team. The tool also encourages exploration by showing multiple design options, which is useful for brainstorming.
I can see how I'll use this repeatedly for quick and dirty designs, while the ability to import finished designs into Figma is a boon for UX designers. Yet here is one of those decisions by the Google team that has me scratching my head: you can only send your files to Figma when using the Standard (Flash) interface, but - bluntly - the designs this produces are underwhelming. I would love to use Pro much more, but with that there's no simple export to Figma option. Likewise, it works best for single screens or websites rather than full mobile apps or sites with many connected pages. The designs it generates are neat but sometimes too simple, particularly in Standard mode, and while this is helpful for a clean interface you still very much need the human touch to add polish. To get the best results, as with most AI apps, you also need to be clear and specific with your instructions as vague prompts lead to generic output. As an experimental tool, it also doesn't link through to Gemini payment plans - which is a real limitation in terms of using the much more impressive Gemini Pro as I would run out of steam on designs very quickly.
As such, Stitch could be a useful part of a web designer's toolkit, but those limitations really do mean that it works better for beginners rather than experienced users. Output is a bit generic although although you will at least have a wireframe to work up in other applications. I'm a great fan of Gemini Canvas for producing specific pages and features, though Stitch seems to lack some of that flair. The restrictions on using the Pro model, however, particularly for transferring your designs to Figma, is the main reason why it's not so useful for professionals.