I Have a Problem — I Write Like AI

I have a problem. I write like AI.

Even when I write completely from scratch, I somehow end up with pieces that score anywhere from 50% to 100% as “AI-generated” on many common AI-writing detectors. I know it’s because I write clearly, concisely, and grammatically correct content, usually targeting a 12th-grade reading level. And apparently, I share with AI a penchant for em-dashes and Oxford commas. Maybe I toss in the occasional clever turn of phrase, but I don’t go overboard. This is exactly what I was taught in high school, college, at work, and even as a published author. Turabian. Chicago Manual of Style. AP Style Guide. Fleisch-Kincaid score. I’ve followed these rules for years, but now it seems like the very skills I’ve mastered are betraying me.

In desperation, I’ve tried using “humanizer” apps, even asked ChatGPT to humanize my text — results were awful. At best, I get something overly casual and clunky; at worst, it borders on LinkedIn broetry. Every time, I’m left thinking, “I could never publish this under my name.” It feels disjointed, like I’m trying to sound human in all the wrong ways — exclamation marks where I’d never use them, strange word choices that clash with my natural style, and an overall tone that feels forced.

But here’s my fear: the backlash against AI-generated writing is going to hurt people like me — good writers who have done nothing wrong, other than produce the kind of clean, structured content that AI can also generate. The irony is almost painful. For years, we’ve been told to aim for clarity, structure, and readability, and now those very things make us sound robotic.

So why does this happen? The problem is that AI has been trained on good writing — our writing. AI models like GPT are fed vast datasets that include high-quality human content. This content follows best practices: clear transitions, logical flow, polished tone. In short, AI writes the way we’ve been trained to write. The detectors, which are supposed to catch AI-generated text, end up flagging anything that aligns with these principles. It’s like our own success in mastering the craft of writing is what’s getting us caught in the AI net.

I see the underlying issue: clarity, coherence, and correctness are now associated with machine-generated work. It’s frustrating to be penalized for writing “too well” because that’s exactly what we’ve been striving for all along. The more polished and professional our work, the more likely it is to be flagged. That’s why when I try those “humanizer” tools, they make my writing worse, not better. They’re trying to inject quirks and casual errors, but that’s not my style — and frankly, it shouldn’t have to be.

There’s an added irony in that the people who judge content by these standards may think that a piece is more “human” if it’s less polished — if it has a few rough edges. But for those of us who have spent years refining our writing, asking us to intentionally degrade our style feels like asking a musician to play out of tune just to prove they’re not using auto-tune.

So, what can people in this situation do? The answer isn’t simple, but I’ve found it helps to focus more on voice, perspective, and injecting a bit more of me into the writing. It’s not about suddenly becoming casual or sloppy, but leaning into the unique ways I approach a topic. Adding personal anecdotes, using unexpected analogies, or breaking a rule here or there for the sake of voice can go a long way. The key is to do this in ways that remain true to my style — no over-the-top informality or awkward phrasing just for the sake of dodging a detection tool.

If you’re in the same boat, I feel your frustration. You’ve honed a skill that’s now being used as a template for AI, and that leaves you at a strange disadvantage.

But I think there’s a way forward — by doubling down on the nuances that make your writing distinctly yours. In a world where AI can mimic structure and coherence, what makes us stand out as human writers is the subtle touch of personality, perspective, and authenticity. It’s about writing in a way that only you can, with quirks and insights that algorithms can’t replicate.

The sad truth is, we might have to adjust to this new landscape. It’s not fair, but neither is the expectation that we should abandon the clarity and skill we’ve worked so hard to develop. The challenge now is to find a balance — remaining polished while also letting our individuality seep into the writing. By focusing on those little things that make us human — the specific way we turn a phrase, the particular rhythm of our sentences — we can push back against the AI tide, even as it keeps rising.

I don’t write like AI — AI writes like me.

P.S.

Justdone thinks this article is 97% AI-generated

Quillbot — 85%

AI-Detector — 78%

GPTZero — 66%

Originality.ai—51%

Scribbr — 20%

AI Humanize — 5%

WriteHuman — 1%

Grammarly and Phrasly (correctly) think it’s 0% AI-generated.

Choose your AI detector carefully. And don’t punish good writers in your visibility algorithm, publication submission, grading, or hiring.

For more on this topic, see:

OpenAI confirms that AI writing detectors don’t work | Ars Technica

Most sites claiming to catch AI-written text fail spectacularly | TechCrunch

Why You Shouldn’t Trust AI Detectors | Lifehacker