AI interfaces and the role of good writing

Why clear, strategic writing is more critical than ever.

An outline of a pencil surrounded by AI icons that look like stars.

If you’ve used an AI product recently, you probably know that the technology is incredible. The UX? Not so much.

Maybe you’ve gotten comfortable writing prompts or using simple one-click tools. But as AI interfaces start to take different forms, many of them are still kinda hard to figure out. Navigating them can be overwhelming. It doesn’t feel like you’re using these products so much as deciphering them. The engineering is powerful, but the flows don’t make sense.

If you’re lost in an AI user flow, blame the writing.

Lots of these tools use unclear labeling, make phony promises, or simply cram interfaces with vague, hyphenated phrases like “AI-enabled.”

The usability issues make sense, because AI is relatively new. We’ve seen this happen before. Think about the first time you tapped on a smartphone. Or the first time you sent an email. Every big shift takes a little getting used to. As new technology arrives, new interaction patterns emerge. And the role of words in the interface needs to be figured out.

The solution: good UX writing. Smart, strategic, honest language that keeps users clear and makes products simpler. It’s more important than ever.

So if you’re designing an AI product, here are some things to consider.

Define your audience

And do it before you write. The first step of writing is understanding your reader. This decision will influence the way you write. It’ll also influence your feature set and product roadmap. Different audiences need different things. An AI fact-checking tool for journalists will look very different than one for biology students.

Three different CTAs with different phrasing speaking to different audiences. — Understand who you’re writing for and how they’ll use your product.

Decide where AI sits in the product

That means shaping the strategy and architecture of whatever you’re building. Before you add AI language and icons onto every screen, align on your approach to AI. Does AI live across the whole product, or just one feature? Is there a character that sits “on top” of the interface?

Now is also a good time to ask: is your product using AI at all? Have an honest discussion. If it’s not AI, don’t say it is.

Three screens that introduce an AI product in different ways. The third doesn’t say “AI” at all. — Consider how you’ll introduce AI, if at all.

Guide the user input

Millions of people are still learning how AI products work. So when you’re writing, think about the first-time user. Will they know what to search? Which file to upload? Where to go for help? Find ways to share guidance or give examples. The best products make sense from the very first screen.

Three different ways of showing a search bar. One screen has no suggestions, the other lists topics, and the third shows actual search queries. — Try different ways to help users get started.

Add structure to the writing

Formatting matters, especially for readability. As you’re choosing good words, think about form, size, color, and spacing. Find the right approach for your information. Maybe that’s a perfect paragraph. Increasingly, it’s short, scannable pieces of text. Try different ways to show the same idea. You can work with the engineering team to create guidelines around how text gets displayed.

Three screens showing how color, formatting, and lists can add clarity to writing. — Explore different ways of displaying the same information.

Choose smart, specific actions

This is one of the toughest pieces of language to land for AI tools. What is the product actually doing? Language shapes understanding, so try different ways to approach the same feature. Play around with phrasing to see what works best. A button that says “use AI” isn’t precise enough.

Three buttons that say different things about how they’ll change an image. Adjust colors, enhance image, or touch up. — Be specific about what your AI product is doing. Photo by Cole Keister

Give a helpful next step

The best AI products are more than intuitive — they’re predictive. They understand user needs and suggest a path forward. The next time you’re designing onescreen of your product, consider what might come next. Explore common use cases and try to avoid dead ends.

Three versions of next steps after a query. If a user asks about an author, maybe they want to visit a library, buy a book, or show the author’s other books. — Use the interface to help create a path forward.

Label your interface honestly

Whatever your product, make it clear what came from AI. Users should know if they’re talking a chatbot—or a person. They should know if the song was written by a musician—or generated by an app. The solution is simple: use words and images to tell the truth.

Three versions of how content can be labeled depending on if it was created by AI or a person. — Be honest about how AI was involved.

Show your sources

It’s a way to build trust with the user. It’s also a way to keep people safe. Showing your source doesn’t feel all that urgent if you’re displaying NBA scoring titles. But AI products are now commonly used for things like nutrition, finance, and medicine. Users should know where important information is coming from.

Three ways an AI-generated response might cite the source material.

Write for all people

Many of today’s AI products were built for early adopters. These products use the language — and acronyms — of engineering. And they presume a deep fluency with technology.

As AI products move into the mainstream, the writing in the interfaces should evolve, too. The writing should become more readable. Respectful. Inclusive.

Crafting that language takes talent and experience. It requires deep expertise and a sense of empathy. Sure, AI tools can generate endless amounts text. But to build and ship great products, you need people on the team who can actually write.

If you liked this article, you might enjoy some other things I’ve written about digital product design:

AI interfaces and the role of good writing was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.

This post first appeared on Read More