“Make it sound human.”
That ideology is everywhere nowadays. But if you sit back and really unpack what you’re being asked to do—or what you’re asking yourself to do—it’s rather silly.
You, a human, ask a machine to write something. Then you take the time to analyze what it generated with a questioning eye and edit it to convince people that you, a human, actually wrote it.
It’s a strange, cyclical conundrum we’ve put ourselves in.
Why not cut out the middleman?
The endless loop
Whether it stems from a lack of ideas, uncertainty about how to start, or an urgent deadline, tasking generative AI to write the first draft seems like a no-brainer. If the technology exists, why not use it? And considering many people are intimidated by writing, thinking it's an innate gift when, in actuality, it’s very much a skill that can be learned, I'm sure that for a majority of professionals, they wholeheartedly believe that the AI’s first draft will be miles better than anything they could ever come up with themselves.
So they don’t question the steps that come after.
Everyone knows at this point that you cannot take what an AI platform generates and use it without some kind of thoughtful audit. The larger public of consumers is apt to spot "AI slop": the too-even rhythm, the same three transition phrases showing up in every other paragraph, the sentence that’s saying something without actually saying anything. And once they deem something as such, you've most likely lost them.
People crave authenticity. They always have. But it's harder to maintain that, or even defend your authenticity, with the proliferation of generative AI.
So you're tasked with editing that first draft. You may be diligent for the very first piece, running it through a real edit, cutting what doesn't sound like you. But over time, you convince yourself that you've given AI enough rules, enough parameters, enough training to write like you after the first shot.
Thoughtful editing becomes a “waste of time”.
You switch up a couple of phrases. Delete the obvious AI lingo.
Then you post.
Taking your mind and your creativity out of the equation to move faster.
And it works, until it doesn't. Until the draft you didn't fully read goes out with a tone that falls flat or a point that was never fully fleshed out. But by then it's not a few minutes of editing you've lost. It's your credibility and the good faith that when you say you wrote it, you actually did.
So why the rush when there’s more at stake?
In theory, it makes sense
You'd think faster is better because time is money. You can completely skip the hardcore brainstorming, outlining, creative thinking, and strategy before phrases and actual words are drummed up.
And it’s a machine, right? It should be spotless: no grammatical errors, a consistent train of thought, clear organization, and structure.
Yes, you may save time and cost upfront. But not in the long run. And yes, AI will probably write a stellar piece that could pass a quick academic read for simple mechanics. But that's about as far as it can go.
Because when you take that draft and reverse-engineer a human voice into it, you inevitably lose what makes writing and language an art form. Creative expression at its height. And the hidden cost is your voice.
How? Because AI doesn't write from a point of view. It writes from an average.
It's pulling from the midline of everything ever written on the topic, smoothing out whatever made any one version of it unique. So instead of the sentence you'd undoubtedly write, you get the sentence that's most likely. Passable but forgettable.
Writers spend their whole lives figuring out who they are through the words they meticulously choose. That's not a small thing to hand over, nor should it be.
Bring back the “old school” approach
So it all boils down to this: just write that first draft yourself!
This isn't a blanket rule for everything your team produces. Informational writing—documentation, straightforward web copy, anything where the job is to convey facts clearly—can genuinely be quicker and cleaner with AI's help from the start. That's not where the problem lives.
The problem lives in creative work. Blog posts, brand copy, anything hoping to carry a point of view. That's where "make it sound human" gets asked of a machine that was never going to sound like a human in the first place.
For that kind of writing, do the major thinking on your own. Flex those muscles in a genuine way that cannot be overtaken by AI. You'll always get a better end product by starting there because you know what you want to say, even if it takes some time to brainstorm exactly what that is. You know how a real human wants to be spoken to and how they’ll respond. Don’t bother faking it or manufacturing it simply because you can. Question if you should.
So maybe the fix isn’t a better prompt for sounding human. Maybe it's not needing that prompt at all.
