Most product teams treat UX copy as a finishing touch. The flow gets built, the screens get designed, and then someone fills in the labels, button text, and error messages at the end. AI has made that habit worse, not better. When a tool can produce copy in seconds, the temptation to hand it the whole job is hard to resist.
But copywritten that way tends to feel like it. Generic, slightly off, missing the specific language your product actually uses. At Reown, we ran into that problem early and decided to take a different approach, one where AI supports the process rather than runs it.
Here's what that looks like in practice, and what we've learned from building it.
Brand Voice and UX Writing: Related, But Not the Same
A common misunderstanding: brand voice and UX writing are often treated as the same discipline. They're not.
UX writing is the functional language inside a product. The labels, errors, empty states, confirmations. The words that help a person understand what's happening and what to do next. Brand voice is the personality layer that rides on top.
Pam Schiavone, WalletConnect's lead product designer, puts it plainly:
“UX writing's first job is clarity and getting the user through the task. Voice and tone is the personality layer that rides on top. Software that responds in a robotic way feels cold even when the flow technically works.”
Both need to work together. Neither can be treated as optional.
What Users Actually Need From Your Copy
Good UX respects the user's time and attention. At every step, users should know three things: what's happening, what to do next, and what happens if they act. If any of those is unclear, the design is pushing work onto the user.
In the digital asset space, this goes further. People need to feel safe before they understand anything else because there's real money on the line. Trust isn't a nice-to-have. It's a prerequisite.
Why Handing UX Copy to AI Can Quietly Hurt Your Product
Handing all UX copy to AI creates a specific and compounding problem. As Pam warns:
"The dangerous part is that plausible but wrong copy now ships at scale."
And beyond inaccurate copy, there's a subtler cost:
"It doesn't know your product, your terminology, and has no idea what the user has at stake at that specific step," she warns. "It reads fine but it's often inaccurate for your product or feature and it quietly erases the product's voice. Everything converges into a generic tone, that over-helpful AI voice. For a product trying to feel distinctive and reliable, that's a real cost."
How We Use AI for UX Copy Without Losing the Human Layer
Rather than using AI to generate copy, Reown built an AI-powered UX Copy Audit Skill trained on Pam's guidelines, brand voice rules, and UX writing decisions accumulated over 14 years of product design experience.
The process works like this:
- Human writers produce UX copy built for its specific purpose
- The AI skill audits that copy against established guidelines and return a rewrite
- A small team of reviewers familiar with Reown and WalletConnect's brand and products checks before anything ships
As Pam describes it:
"Use it to set a standard, not to replace judgment. The best use is as a second opinion based on your own rules: your guidelines, your glossary, your decisions. It makes the team's tacit knowledge explicit and applies it the same way every time, regardless of who's reviewing."
The AI enforces the standard. The person owns the judgment.
Three Things We Learned After Three Weeks in Production
A few weeks of production use surfaced three meaningful lessons.
The cultural shift came first. The whole team now understands that UX copy is not trivial. It stopped being the thing you fill in at the end. As Pam puts it, copy is now treated with the same care as using the right colours or the right assets from a design library.
Visibility improved. Every audit ends with a ready-to-paste handoff to a shared channel. Reviews don't die in a DM. The team can see what's actively being worked on and learn from each other's examples.
Standardisation revealed what to improve. Three weeks of real use surfaced gaps in the guidelines themselves, things to add, rules to sharpen. And it made one thing clear: if you want an optimal result, you still need the human eye. UX writing is not something you can do 100% with AI.
Six Things to Get Right When Writing UX Copy
For teams building early-stage products, Reown's approach comes down to six principles:
Start with the personality, not the screens. Decide who the product is before designing anything. Every later decision then has something to measure against.
Give the user context. At every step they should know what's happening, what to do next, and what happens if they act. Don't forget trust if money or anything sensitive is involved.
Write the copy with the design, not after it. Copy exposes where a flow isn't clear. If you can't write the screen in a few honest words, the screen isn't finished.
Understand what "simple" really means. Simple doesn't mean cutting copy. It means doing the hard work of deciding for the user so they don't have to carry that burden.
Start a tiny glossary before you think you need one. One word, one meaning, used everywhere. It's painful to untangle once you've scaled.
Don't copy a pattern just because the big products use it. Much of what looks like the standard is inherited, often a technical constraint everyone is repeating. Starting fresh means you don't carry that baggage.
The Part That Has to Stay Human
The standard can be automated. The judgment cannot.
Copy is the voice of a product. It opens a conversation. Nothing replaces what a human catches with their own eyes; the pauses registered in a user interview, the questions that come to mind in the moment. That part has to stay as human a conversation as possible.
Democratize the draft. Own the review.
Want to go deeper? Read the full article on the Reown blog, including Pam's complete take on product personality, the dangers of AI-generated copy at scale, and how to build a UX writing process that scales without losing its voice.