Why does every AI product look the same?
Seriously — open any AI-powered tool launched in the last two years, and there it is: a text box, a blinking cursor, and an invitation to "ask me anything." We're designers. We've spent decades refining how humans interact with software. And somehow, we've collectively agreed that the most transformative technology of our era should ship behind the digital equivalent of a blank page.
That's a design failure. And it's one we should be embarrassed about.
What happened
There's a growing conversation in the design community about breaking free from the chat-prompt paradigm that dominates AI interfaces. The argument is straightforward: the text box has become the default interaction model for AI assistants, and that default is constraining what these tools can actually do.
The chat interface made sense as a starting point. Large language models are, by nature, text-in-text-out systems. But we've confused a technical constraint with a design decision — and then stopped thinking about it entirely.
Meanwhile, the consequences of lazy defaults are showing up elsewhere. In markets like Singapore, AI-generated advertising is flooding in at speed, and poorly executed work is actively damaging brands. When you combine a low-friction creation tool with a low-effort interface, you get a lot of mediocre output very quickly. That's not an AI problem. That's a design problem.
The bigger picture
The chat box has become the hamburger menu of AI — a pattern so ubiquitous that nobody questions whether it's actually good.
But think about it from a designer's perspective. When you're working visually — adjusting layouts, exploring colour palettes, refining spacing — the last thing you want is to describe what you're doing in words. The whole premise of visual design is that some things communicate better through space, colour, and gesture than through language.
Yet we keep asking designers (and users) to translate their spatial, visual, intuitive intentions into typed sentences. That's backwards.
The tools that are getting this right are the ones embedding AI directly into the workflow rather than bolting on a chat panel. Think about generative fill in image editors — you select an area, and the AI understands from context what should go there. No prompt required. Or consider AI-assisted layout tools that suggest adjustments based on what you're already doing, not what you type into a sidebar.
These aren't just nicer interactions. They're fundamentally different design paradigms. They treat AI as a collaborator that watches and responds, rather than an oracle you have to petition with the right words.
And here's what I think most people are missing: the chat-box default isn't just bad for users — it's actively limiting what AI can become. When the only input is text, the AI can only respond to what you can articulate. But so much of design is about the things you can't quite put into words yet. The feeling that something's slightly off. The instinct that a layout needs more breathing room. A good AI interface should meet you in that ambiguity, not force you to resolve it before you've even started.