Writing under AI surveillance: how humans are flattening their own voices to avoid being flagged

Even when no artificial intelligence (AI) is watching, writers are rewriting themselves, flattening their voices, policing their words, and second-guessing every sentence as if an invisible algorithm is judging them.

I was writing a student report for a teenage English student of mine the other day, and I felt like I was being observed. Even though there were no parents, supervisor, or AI tools in sight.

Whatever I wrote felt watered down, and this was far from being an overly expressive text. “Sebastian has had a good year,” I twitched, whatever that meant. It felt like someone else was in control of my fingers, as if algorithmic content has had a cerebral effect somehow.

When I get notifications now for LinkedIn posts, I instantly flag it as slop without batting an eyelid. But who am I to knock something before I try?

The thing is I have tried. The exception is when I open a weekly newsletter like the Cybernews one. Biasedly, I enjoy reading this, as the human writer often spins a personal yarn about something weird that’s happened to them (or in the world) that particular week. In other words, there’s a sense of time and place.

But how about when there isn’t? Writers now seem to call each other out and over-police one another over em dashes. And aside from that, even when writing a bare-bones student report, it feels like someone is peering in.

Joshua Lisec, a New York Times bestselling co-author and internationally acclaimed ghostwriter, has spent years coaching writers on style, persuasion, and tone. He says that AI surveillance is already reshaping how humans write, whether that’s fully, or half-half:

You can see where someone has stitched together, for example, their own original thoughts with AI, where they'll kind of ramble for a little bit, and it won't make any sense, and then suddenly it starts making sense. And it doesn't just start making sense – it provides logic, reason, understanding.

Though inconsistency within a piece can be more of a subtle giveaway, there are some AI hallmarks that are as clear as day. For starters, the overuse of colons and em dashes is one giveaway, but writers can still remove them and still pass off AI’s content as their own work.

And Lisec points to political speeches and YouTube videos being instantly dismissible when they use a particular rhetorical structure that screams AI:

“The ‘not this but that’ or ‘not just this but that’ structure. And it's not just about using AI – it's about using AI effectively.”

No shame whatsoever

I just opened my LinkedIn, and one defender of AI-written content included: “So maybe the real question isn’t 'Who wrote this?' It’s simply: 'Did this move me? What about you? If you found out your favorite book was co-authored by AI, would you love the story less?”

I’ll spare the writer the attention, and not name them – and of course I would love the story less, as I believe we should consider the source. It wasn’t even a difficult one to answer, probably because he didn’t write it anyway.

And as his post came back as 100% AI-generated, the writer's objective is exposure more than anything – the cheapest click matters.

Writing has shifted from expression to exposure. Text feels like evidence of behavior, and authorship feels contestable, even in private drafts.

That feels precisely what the LinkedIn post is doing, causing provocation, but using AI to do it in a matter of seconds.

Walking on eggshells

When we scroll content nowadays, it seems like we’re red-flagging content rather than actually absorbing it.

But even when a writer is for real, the subliminal effect that AI has had on us is astounding. Just as a summer study shows that people speak like ChatGPT in everyday chatter, with writing, it can constantly feel like you’re walking on eggshells.

Even when no AI tool is being actively used, writers anticipate algorithmic judgment. That awareness alone changes how people phrase sentences, choose words, and structure paragraphs.

“In my case, and in many people's cases, we're using the em dash less and eliminating 'not X but Y' from our writing, even though those are things that we’ve used for years, simply because we don’t want to sound like – or rather read like – we're using AI,” highlighted Lisec.

I’ve also come to notice a clipped style in a writer's vernacular. Swathes of “no-fluff” guides of straight-talk. Paradoxically, writers love the word “niche” when it’s become the go-to adjective. And the brazen sounds much less brazen with the overuse of “here’s the brutal truth about…”

AI policing AI

We’re in a looping state of mind, at present. We use AI to police AI and like to blurt out artificial responses and pass them off as our own.

And as output is crunched, whirred, and churned out at an industrial level, the feedback loop is in place as the algorithmic content is more baked in than an encrusted potato in the oven.

It's recursive. The AI is reviewing itself at so many of these large accounts. But they're not likely to be banned or get into trouble because the AI output probably won’t produce anything that gets banned.

Joshua Lisec, International Bestselling Ghostwriter

With machines exempt and humans reluctant to put themselves on the line, unless the AI bubble bursts one day, we’re writing slop in spite of AI, not because of it.

“So now, the only writers on the internet who are at risk of being censored are humans, by AI. Isn't that horrible?,” Lisec pondered.

Unlock more exclusive Cybernews content on YouTube.