Who Are You Writing For? Readers or Detectors?

It's getting increasingly hard to detect what's human and what's not, what is written by someone and what is written by no one.

Recently, I received a comment under one of my substack posts that read: "Girl I'm scared. Why does your writing have these absurd hyphens. AI uses them."

First off, I'm pretty sure they mistook Em dashes for hyphens. And as much as I'd hate to admit it, that comment got to me. For the first time, I found myself questioning my writing, wondering if a punctuation mark was a hill worth dying on. But the problem isn't my choice of punctuation. The real question is: when did something as little as an Em dash become a telltale sign of AI writing?

That comment wasn't my first AI allegation. A few months ago, I ran some substack articles through AI detectors and ended up writing a whole piece about how LLMs and AI detectors are keeping us in an ouroboros loop of good cop/bad cop.

First, AI was introduced to help make our work easier and now detectors are being used as judges to penalize writers for using said tools at their disposal.

In an era where well-written prose with clean grammar, structure and pacing is enough to make the feds come banging on your door at three in the morning, what does this mean for writers?

Some writers are now deliberately writing worse—breaking flow, avoiding the em dash (defend the em dash), auditing their voice, adding mistakes dressed up as "stylistic imperfections" to avoid being accused of cheating. And I'm not the only one doing it. Several readers said the same under my recent substack post. A writer on Minding The Campus admitted to dropping Em dashes altogether. The same writer accused a student of using AI to write her essay because it was too polished. Not long ago, a well-crafted essay would have been met with praise and admiration, but today, it's met with suspicion and scrutiny.

This gradual decay of trust between writers and readers has caused people to adopt the "Fake until proven real" mentality. The real question shifts from: "Who's the engineer behind this idea?" "Whose own lived experience shaped their writing?" "Did I enjoy this piece?" To "Did AI write this?"

Subconsciously, writers are learning to view their writing from the cold lens of a scanner. We stop asking: "Does this make sense?" "Does my reader understand me?" And start wondering: "Do I sound human enough?"

Sometimes, I find myself in the same clusterfuck—staring at my blinking cursor for what feels like hours, rereading the same paragraph over and over again, wondering if every word sounds "too textbook," too smooth, too polished. And because I can't prove the smoothness is mine, I start sanding it down (aggressively). I stop writing, and I start fixing. Fixing my voice—adding slang words I never would’ve used before, deliberate typos, wrong punctuation, rewriting every sentence until I can't remember where my thoughts begin and where my words end—each revision isn't "just a revision" when it erodes the jagged planes of our personality, leaving only the hollow, bleached bones of our once-vibrant thoughts.

This is the part that frightens me the most: If we have to audit our voice just to sound more human doesn't that mean writing becomes inauthentic?

Writing in itself is performative. We exaggerate some cues for flair (or a good laugh). We downplay a scene to match dialogue. We create imaginary scenarios to convey a message to the reader (I do it sometimes!).

But when writers are scrambling to sound as "human" as possible, the real human element—the natural quality of their writing—is lost, and the performative quality of writing is at risk of veering into the inauthentic.

This dance of adapting one's style to avoid accusations of AI use creates a loop, because once one problem is solved, it gives way to another. Most AI detectors are unreliable, so even experienced writers get flagged mistakenly. In response, they dumb down their writing, and a flattened prose becomes the standard everyone tries to emulate.

If we read the same bland AI-written texts, subconsciously, we learn to mimic the same writing styles: similar wording, structure, pacing and rhythm. Writers sand themselves down out of fear, and everyone reads that sanded down version and absorbs it.

It's ironic because LLMs are trained on human writing, which inadvertently means they evolve alongside us—they learn to mimic our styles, polished or not.

The fear of sounding like AI will create a new kind of conformity, producing the exact same repetitive writing we blame robots for.

The real threat we writers face isn't a worse sentence, it's the missing (or cosplayed) personality. And the fix isn't to sand down your writing, it's to be more specific—write the things only you notice, share memories only you have experienced, share opinions that you commit to.

I often tie a personal experience to my writing, because that's my own way of telling readers that a real person wrote this, not a chatbot.

Stop treating every essay like a test and the detector as your strict grader. You can't out-imperfect a machine that is built on your creativity, and you can't escape the performance either, but you can choose who you're performing for.

I'd rather receive a thousand allegations for being true to my writing than be praised as a human for writing like everyone else, because whatever makes writing human lives in the uniqueness of the heart and the specificity of the mind behind the words.

SOHBAÉ is a critical overthinker who writes about writing, self-reflection, and technology. She is also a diehard Zara Larsson fan. Read more of her work on Substack.