Hi there,
Wow — it’s been a while since I sent out this newsletter! How have you been?
First, some updates from me:
I just launched Supershort — a UX writing and content design studio that makes complex products simple. We’ve worked with Spotify, Minecraft, WordPress, BeReal, Miro, and many other companies. Need senior content design help for a sprint, audit, or ongoing support? Reply to this email and let’s talk!
I also created Vibesuite, a Chrome extension that makes vibe coding tools like Bolt, Lovable, and v0 easier to use. Try it for free!
I'm planning a new and exciting direction for this newsletter, and you’ll hear more about it soon.
Enough housekeeping. Let’s talk about how AI is changing content design.
— Juan (follow me on X and LinkedIn)
Over the last couple of years, AI quietly moved from a fun tool to being indispensable at work. Instead of staring at empty text layers in Figma, a first pass can now appear with headlines that aren’t terrible, buttons that mostly make sense, even empty states that sound like they’ve met a human before.
That’s great for momentum. It also means we’re suddenly swimming in options. As UX writers and content designers, our job shifts from writing everything to deciding what deserves to live. To make that work, we need tighter rules, better structure, and quick ways to check whether our words actually help people do the thing.
AI drafts, we decide. Speed is free. Clarity isn’t.
It’s important to note that being AI-first is now often a leadership stance. Shopify publicly told teams to default to AI and prove it can’t do the job before hiring. For UX writing, that means our job shifts from typing to deciding what ships and owning the guardrails — voice, consent, clarity.
What does this mean for UX writers and Content Designers?
A lot of people have asked me the same question: is this job done for, or does it open new doors?
My take is that it opens doors... if we lean into the parts humans can do well.
Less blank-page panic. A “good enough” draft is easy to spin up, so we can spend more time on intent and hierarchy than on filling placeholders.
Scale by default. Variants, locales, and states can be generated together. That sounds dreamy until you’re juggling fifteen similar strings — this is where terminology, voice, and structure matter.
Context still wins. Models miss edge cases, hierarchy, and brand nuance. They don’t know your consent policy or the places where trust breaks. You do.
Avoid sameness. AI trends to the median; we can keep things fresh by safeguarding output.
Tools are improving fast, but human judgment is still the safety rail.
Okay... so how do we use it?
AI gets better every month — just look at the breakneck pace at which new models are coming out. The best way to get ready is to play with it on real work and keep the parts that actually help.
ChatGPT, Claude, Gemini — used well, they’re great creative partners for product writing. Used lazily, they’re a liability — and remember: don’t paste in sensitive data that you wouldn't want a big AI company to see.
How I like to use it
Structure — “Here’s the flow. List the states I’m forgetting.”
Voice check — “Using these rules, mark anything off-brand and propose fixes that fit the space.”
Options, not answers — “Give me 3 takes with a one-line why. No fluff.”
Recent surveys show that more and more teams are using AI beyond drafting — templating flows, checking tone, summarizing notes, and generating variants at scale. If you don’t have a stable workflow up yet, it’s time to do so.
How I don’t use it
I don’t paste outputs straight into a product. AI is very good at making confident mistakes and still struggles with context and hierarchy. Treat it like a sharp intern: helpful and fast, but needs supervision.
Experiment with a custom GPT
A quick win is creating a small, purpose-built UX writing GPT for your team. Feed it your voice rules, glossary, and do/don’t list, and point it at a few real screens. Now you’ve got a lightweight helper that can provide support and suggest fixes without rewriting everything in robot voice.
Quick setup (ChatGPT):
In ChatGPT, go to Explore GPTs → Create.
Name it (e.g., “Content Design Helper”) and write clear Instructions (what to do, boundaries, and format). Example: “Return 3 options with a one‑line why; never change product terms; flag off‑brand strings under 50 chars.”
Under Knowledge, upload your voice guide, glossary, do/don’t list, and 3-5 good/bad examples.
In Capabilities, enable only what you need; leave browsing/code off unless required.
Save and share as Link‑only with your team.
This way, you have a fast, lightweight AI helper that can quickly give you (and your team) feedback on any work that’s being shipped.
What you can do right now
With your new custom GPT helping and you deciding, here’s a small loop you can integrate in your workflow without slowing down your current processes.
One AI-assisted rewrite on a real screen.
Pick something with constraints (space, states, real data). Generate a few options. Keep the useful bits, rewrite the rest in your voice, and make sure the hierarchy still guides the eye.
Tiny test, zero drama.
Show original vs. final to three people on your team. Ask: Which is clearer? Where did you hesitate? Ship the winner and turn what worked into a rule.
Save the pattern.
If a line or pattern keeps winning, document it once so you don’t have to rediscover it next time.
Final checks
Finish up with a short checklist before anything leaves Figma — facts and names are correct, tone matches the moment (especially errors and consent), accessibility is ensured, and localization won’t break anything.
UX writing in the wild
As any UX writer knows, it’s important to make sure you’re writing with localization in mind. Even if it needs to be translated to pirate.
That’s it for this issue — thanks for reading all the way to the end! Let me know what you think by replying to this email or sending me a DM on Twitter.