Editorial Thinking for GenAI or: How I Learned to Stop Worrying and Love My GPT

Note: This post contains a lot of em dashes. So does everything I've written since the late '90s. This does not mean I'm an LLM.

From the Editor’s Desk to AI Input Fields

My first job after college was as an editor at a pre-Internet music magazine. When the Web exploded and content farms cheapened what I'd once held as high art, print began to feel like a dead end. But the fundamentals I’d learned as an editor — managing workflows, crafting narratives, and shaping language for clarity and impact — didn’t lose relevance. They paved the way for a career in content design, where I’ve spent the past 15-odd years running from obsolescence.

Generative AI’s impact on UX content feels eerily similar to the way BuzzFeed disrupted print, down to the irreversible human and workforce impacts. I don’t want to dismiss these. But as a human with bills to pay who’s deliberately chosen to participate in late-stage capitalism, I’m aware there’s no putting the AI genie back into the bottle. So I’ve jumped headlong into vibe-designing and experimenting with the tech. And through that experimentation, I’ve learned some things.

First, GenAI is not an extinction-level event for content designers, even if it has devalued those who see themselves solely as writers without broader UX fluency. Second, it has dramatically increased the need for content designers who can work beyond words to solve problems of strategy, systems, and structure. Third, it enables content designers to scale our thinking and streamline mundane tasks, which is always helpful for job security.

So, fellow content designers, here’s how I’ve applied my editorial foundations to stay relevant (and even thrive) in the age of AI.

1. Think Like an Editor

Magazine editors think in concepts and big pictures. They set themes, forecast trends, and shape narrative arcs long before assigning stories to writers. They plan with the end in mind, then they assign, revise, and shape each writer’s output to serve the whole.

Generative AI, in this context, is the freelancer I always wanted. No ego to massage, no flaking on deadlines. When I create prompts for ChatGPT or Claude, I imagine I’m briefing a writer. I provide purpose, context, structure, and constraints. And, yes, I bring “taste” and “craft” (even if I think those words have been co-opted by design snobs and white dudes in power positions). Let's call it the sum of my lived experience in writing, editing, linguistics, and human-centered design. The point is, you need foundational chops and critical-thinking skills if you're outsourcing creative work to a robot.

2. Get Clear on the Assignment

LLMs are context-sensitive, non-deterministic, and incredibly bad at vague multi-tasking — just like us! They perform best when you give them clear, well-structured objectives and creative constraints. That means answering some basic questions before you start plugging in prompts:

• Who’s the audience?

• What do they need from the content?

• What’s the reading level?

• Where will the content live?

• What’s the character count or space constraint?

• What tone, voice, or brand system are you working in?

And so on.

All of these are things you already manage as a UX writer and questions you’ve asked if you’ve ever worked as a content strategist. Putting them down on paper helps you think clearly and make sense of the mess you’re in, to paraphrase the great Abby Covert. It also transforms the chaos of stakeholder input into something coherent: a brief. And what designer doesn’t love a good brief?

3. Apply the Right Framework

Prompts are only as good as the frameworks behind them, which are only as good as the language of the LLM instructions. Without structure, you’ll get lifeless garbage, and your first experience with GenAI might feel like a waste of time. But once you define the proper framework for your task, everything changes.

Some of my happiest experiences with GenAI have been the light-bulb moments that happen between content designers (experts in brand and product language) and machine learning engineers (experts in model development and structured data). In one project, a content designer new to model design was peer-coding with an ML engineer, trying to dial in the tone of voice for an LLM with fairly restrictive guardrails. Instinctively, given their experience in branding, the content designer started articulating a voice persona, breaking it down in a structured way the engineer could grok. When the eureka moment came and the model started talking back as the persona, I wish I could have bottled the glee in the engineer's voice.

I won’t catalog every AI prompting trend you could chase, but if you have access to an ML engineer, grab them a coffee and ask them to walk you through prompting and evaluation frameworks. If you're vibe-designing, a simple Google search will open a world of possibilities. Start with one framework, experiment with others, and apply what fits your context. Ultimately, let the situation determine the framework you choose.

I’d also recommend collecting these frameworks into a toolkit for content design work, just as you'd do with Figma templates or messaging architectures. For example, I built task-specific digital twins of myself because I was tired of reinventing the wheel for high-effort, low-value activities. It took significant upfront work, but the investment's paid off when I'm stretched and need to offload the heavy lifting.

4. Accelerate the Crappy First Draft

Even if you’ve been doing this work since UX writing became a thing (hi, yes), your first drafts are still going to be crap. And you’ll face stretches of writer’s block, triggering impostor syndrome when you’re minutes away from a standup and have just one half-written string to share. This is all part of the work we do, but who wants to experience it willingly?

GenAI is great at getting you past the daunting blank page. It doesn’t get tired. It doesn’t take meetings. It doesn’t struggle with impostor syndrome. (Insert hot take here about how maybe it should.)

Need 10 strings for a multivariate test? Ask your GPT for 50. Or 100. Most of those first drafts will be bad. But editing slightly crappy content is often easier than achieving perfection from scratch, especially when your brain is fried from back-to-back context switching.

So let the machine get you started. That’s its job.

5. Show Your Hand

Recently, I jumped in on a few projects that lacked dedicated content design support. I toggled between IC work and my role as a design leader, shipping content that met the team's high standards. And I used GenAI for nearly all of it.

Not blindly, of course. I used a custom GPT that I'd trained on decades of my writing, so the first drafts would be as close to my thinking as possible. Because I’d already built the prompting and evaluation frameworks, most of this took minutes.

I refined the outputs, stripped out the AI tells, and, well... the work still looked like crap. Which wasn't surprising, given that it was my first real-world experiment with the thing I'd built. (This is why we prototype and refine.) But seeing the AI's wonky first drafts did something unexpected: It reignited my burned-out UX writer's brain.

My uncanny twin's output prompted me to think like an editor, which led me to think laterally and explore different approaches to the interface copy. Ironically, it helped me close my GPT window to knock out a handful of variants that tested well and ultimately made it into the app. This, by the way, also took minutes — but it could’ve taken hours without a kickstart from my GPT.

So if you use GenAI as a creative catalyst, invite teammates into your journey. Share your toolbox with others who are facing pressure to deliver under duress. And keep experimenting with those vibe designs. GenAI becomes more powerful when we treat it not as a replacement, but as a partner in the messy process of thinking, shaping, and solving.

Even if we're heading toward the singularity, that kind of self-awareness is still something only a human can bring to the work.

What content design workflows have you adapted for GenAI? What’s working — or not working — for you? Let’s compare notes.