This year, we've witnessed a momentous shift in the creative staffing space. For the first time, I’ve heard leaders asking not who they should hire, but whether they need a person at all.
If that question becomes the new norm, we're in the midst of a bigger work shift than the conversations give it credit.
So, when my long-time copywriter moved on, I didn’t replace her right away. Instead, I put the AI hype to the test. What does it really look like when companies try to replace a living, breathing human—someone with years of experience, training, and individuality—with an LLM? Does ChatGPT really fill the gap? Could a tool actually help us create thoughtful, strategic content for a discerning creative audience?
Everyone was saying the same thing: “AI can write in minutes! It literally knows the entire internet!” But the reality was messier. Slower. More frustrating. And sometimes… a little unsettling.
Here’s what actually happened.
"It made up an entire study that we did. It completely invented a statistic from Artisan Talent."
Writing for Artisan involves speaking to two different, but often overlapping audiences: creative leaders/hiring managers and job seekers. For this experiment, I needed to produce two full-length blog posts that:
Demonstrate that Artisan is a leader in this space and that we understand the needs of our community
Share best practices for hiring and getting hired
Hit relevant SEO targets
Matched our brand’s tone and voice
Included real, not made-up, links
Delivered actual insight—not fluff
Quick note: Previously, I used ChatGPT to copyedit and incorporate SEO. I'd been using it regularly for approximately 6 months, so it already had a strong context for our brand.
When I first jumped into this experiment, my prompts were thin. I quickly learned the value of including a more detailed and thorough prompt. Everything from background context, audience notes, SEO goals, internal links, the articles we wanted to reference, and even reminders of things we'd talked about before.
I even had ChatGPT suggest the two topics we'd write together:
How to Build a Creative Team Without Burning Out (for hiring managers)
Why Your Creative Portfolio Is Getting Ignored (for job seekers)
Both started with promise. But neither ended where I hoped.
The first post took more than 16 hours to get right. That’s two full workdays.
When working with my freelance copywriter, I normally spent a max of 2 hours editing... a simple last-minute read, adding backlinks, SEO keywords, voice/tone tweaks, and crafting the surrounding social media content. She traditionally spent ~5 hours writing. So altogether, I was used to this process taking 7 hours MAX.
The biggest issues:
Hallucinated backlinks: The AI invented articles from our blog that don’t exist.
Audience confusion: It repeatedly wrote to creatives, instead of the hiring managers we were targeting.
Misused job titles: It didn’t understand the difference between a Visual Designer and a UX Lead.
Tone issues: Everything sounded like a content farm—overly neutral, missing edge and insight.
Nonreputable sources: It often pulled information from content across the internet that I didn't think our audience would respect. I ended up having to pull a list of sources I'd like it to stick to.
By the end, it wasn’t a polish job. It was a rewrite.
"AI might replace a copywriter who doesn't deep dive. It is not going to replace someone who's really going to try to understand your business."
The second post went faster—maybe 10 hours total. Partly because I knew how to prompt better. But mostly because it was a much more straightforward topic.
What worked better:
It helped structure an outline more clearly.
The research was much more usable.
It helped me avoid blank-page paralysis.
What still didn’t work:
Tone was still off. Still too general, too safe.
It continued to struggle with specificity.
Examples and phrasing often missed the mark.
Jokes just didn't make sense.
Even with better prompts, I spent more time editing than I wanted to. Just… editing differently.
"It helped me get unstuck. It kept me going through this process. It actually helped me organize some of my messier thoughts."
In the end, AI didn’t write my blog for me. But it did help me:
Clarify early structure and flow
Spot patterns in our strongest-performing content
Stay on task when my energy dipped
Work more iteratively
Understand Generative SEO/LLM Optimization and start to incorporate best practices for this new type of SEO
Think of it like a junior strategist. Not a lead writer.
The three biggest pain points:
Nuance: It couldn’t capture subtle tone shifts or context-specific voice.
Specificity: Analogies, examples, and references were often generic.
Accuracy: It confidently fabricated internal links and even statistics.
I spent as much time fact-checking AI content as I would have spent writing it myself. Also, the writing itself? Even after multiple edits, it still felt a little flat.
"Really struggled with three things: nuance, specificity, and my audience expectations."
Some of the best lessons from this process came down to how I prompted. Turns out, good prompting isn’t a cute trick. It’s a skillset.
Here’s what I learned:
Front-load everything. AI works better when it has all the information from the beginning. The more iterative and fragmented my approach, the worse the output got. You only get one shot at strong context before it starts making assumptions.
Feed it background and structure. Treat your prompt like a creative brief you're passing to a freelancer who is just being introduced to your brand. Before each post, I gave ChatGPT our previous blog examples, our brand tone, audience details, SEO goals, and internal content it should reference. The more it had, the more helpful it became.
Content Marketing Institute also recommends starting with a fully developed outline before expecting meaningful paragraphs.
Always ask for sources and gut-check them. Then do it again. I clicked every link. I searched the stats. I rewrote the parts that didn’t add up. Generative AI tools like to please. And that means they’ll confidently give you wrong answers unless you push back. Harvard Business Review explains this in the context of AI "hallucinations."
Assume it’s overconfident. That’s not a bug—it’s the system working as designed. Which means your job isn’t just writing. It’s interpreting, validating, and re-shaping what you get. MIT Technology Review offers a deep dive on how overconfidence happens.
If you’re not willing to do that? AI’s not going to save you any time.
AI is not a cure-all or a demon. It’s not coming for your job (at least not yet). And it’s not magic. But used well, it can make you faster. More organized. Less blocked.
"Did AI earn the job of replacing my copywriter? Not yet."
But it did remind me what great writers bring: context, courage, clarity, and taste.
Will I keep using AI? Yes. But only in the right roles.
AI is a decent SEO consultant. A passable editor. A solid brainstorm partner. But a writer? Not yet.
“It helped me organize my thoughts, keep the process moving, and not only rely on my own drive to get something done.”
That alone is worth something. Especially in a time when our creative energy is being pulled in more directions than ever.
But if you care about telling the right story, in the right voice, to the right audience, AI still needs a lot of help. And that means the job isn’t going away. It’s just evolving.