A Year in Daegu: An Honest Review of AI Copywriting Tools

Two in the morning. I closed my monitor after checking client feedback in a studio apartment near Dongseongno. "The tone is too mechanical." That was the comment on copy I'd drafted and refined using an AI tool. From that day on, I started tearing apart my workflow, one piece at a time. A year later, I can tell the difference between the tools that actually made me money and the ones that just ate up my time.

The Tools That Survived

ChatGPT ended up being the tool I used the longest. It proved its worth not as a drafting tool but as a brainstorming partner. When I asked it to generate 20 performance marketing headlines, I'd throw out 18—but the remaining 2 would point me in the right direction. The monthly subscription felt like a stretch at times, but one client project was enough to cover the cost, so I kept it.

Claude became my main tool from the midway point on. When I dropped in an entire brand guideline and said "rewrite this to match the tone," the results were more consistent than ChatGPT's. The difference was especially clear with B2B SaaS copy. The deciding factor was being able to feed a 40-page brand book from a client straight in as context.

I used Jasper fairly often during the first three months. The templates made it fast, but the problem was Korean-language quality. English copy was decent; Korean output always needed heavy editing. I was paying around 70,000–80,000 won a month and using it exclusively for English copy, but since most of my clients were domestic companies, I naturally let it go.

The Quiet Breakups

Copy.ai didn't last two months. I had high hopes for short-form content, but every sentence followed the same pattern. It felt like it was just recycling CTAs along the lines of "Do ~ right now." Writing it myself was faster.

Writesonic got cut for similar reasons. I tested it for SEO blog drafts, but the generated articles kept repeating a structure that felt all too familiar. Subheading-body-subheading-body on mechanical repeat—done.

Notion AI is more of a feature than a tool. I only hit the button occasionally to summarize meeting notes while managing projects. I never intended to use it for copywriting, and in practice, it's weak for that purpose.

What Mattered More Than the Tools

Looking back on the year, it was prompt design, not the tools, that determined skill level. Even with the same ChatGPT, there's a world of difference between "write me some copy" and "Instagram ad headline under 130 characters, targeting women aged 25–34, problem-awareness stage." The day I sat in the Starbucks at Banwoldang with my laptop open, just organizing prompt templates—that turned out to be my most productive workday.

One more thing. This past year is also when I started thinking seriously about AI ethics. A client once asked me directly, "Did you write this copy with AI?" I was honest. I told them I draft with AI and then rewrite everything myself. That client appreciated the transparency and kept giving me work. The moment you start hiding it, that's when it becomes a problem.

Right now my stack is Claude, ChatGPT, and occasionally Grammarly. That's it. Having a lot of tools isn't a competitive edge—using fewer tools but going deeper with them turned out to be the right call. I don't see this combination changing next quarter either.