The Single Most Prevalent AI Writing Tell (Spoiler Alert—It’s Not Em Dashes)

I know what you're thinking . . .

Oh, for Pete's sake. Another regurgitated bull about signs something is written by or with A.I. Give me an effing break!

But hear me out.

I'm writing this, at least in some measure, to exonerate em dashes—my and fellow writers' favorite punctuation mark—after being indicted in the court of internet discourse as the ULTIMATE giveaway of AI writer involvement.

(People even have the gall to call it the ChatGPT hyphen.)

Are Em Dashes a Sign of AI Authorship? No—and Yes—Kind of. My response to the allegations as both an AI promptstitute by profession and an em dash addict by passion

Such an accusation has since led to the poor symbol's permanent ban in the custom GPT instructions of AI prompters who don't want their AI content to be suspected as, well, AI content.

While I don't deny—I can attest to it, even—that ChatGPT loves it some em dashes here and there—sometimes everywhere—I've noticed a more prevalent and consistent pattern in LLM writing.

It's a rhetorical technique that Chat and similar tools have habitually been pulling from their artificial drawers when generating effortless, predictable responses.

So much so that I proclaim it as the more telling sign someone used AI to write that "thought leadership" garbage on the AI-prose cesspool that is LinkedIn.

I'm positive I have used said technique in one of my past writings. But I have yet to see other hooman writers use it to the obsessive-compulsive extent that LLMs do when pumping out content on virtually any subject.

The single most obvious AI writing tell

You might guess that the most obvious AI sign I'm going to reveal is indeed the most obvious one.

That is, when the text includes a ChatGPT follow-up artifact that shouldn't have been a part of the copy-pasted information, if you want to pass it off as written by a human.

Like this:

…Would you like me to add a concluding paragraph or perhaps a bulleted summary? Just say the word!

This goes without saying, don't you think? So no, this is not the main reveal.

Hang on a second.

But if I may, allow me to briefly pontificate to click-happy prompt engineers who do an atrocious job at [not] editing AI-generated slop:

I'm begging you to read and review the whole dang text for any claims and potential hallucinations, before hitting publish.

And don't make a fool of yourself by accidentally Ctrl+C-ing the last part where ChatGPT is just offering suggestions.

Save readers from the secondhand embarrassment of reading it and suddenly fighting the urge to chastise you for this editorial negligence.

Thank you very much.

Without more ado, here's what I think is the most prevalent AI tell:

Negation.

If the text uses this tactic to drive a point home, pardon my judgmental ass but I'd say the probabilities are statistically high it's ChatGPT'd.

"[Topic] isn't just / is more than just about [basic definition and understanding of it]; it's about [supposedly deeper definition and interpretation of it]."

Turns out AI writers trained on human copy looooove giving meaning to a subject by first stating what it isn't just about.

Add quotation marks and the rule of threes to the mix—and BOOM! You've just hit the trifecta of a raw and unedited AI-generated text.

I found this contrastive writing structure especially ubiquitous on social media.

As someone chronically online (aren't we all?), I have come across multitudes of content creators who post obviously AI-aided micro-essays on subjects they seem to care deeply about, but lack the fluency or breadth of vocabulary to articulate their thoughts clearly, convincingly, and without inadvertently drawing the attention of the grammar nazis.

As I see it, resorting to advanced software to write better is how they find the courage to "announce" their thoughts online for internet strangers to like and comment affirmatively on.

And for that, at least for me personally, they get a pass.

For dramatic and poetic effect

In all fairness, negation as a rhetorical device works pretty well in terms of quotability.

It creates a dramatic, even poetic effect and resonates well with today's digital doomscrollers. That many of these AI-generated posts even get traction and/or achieve virality is a case in point.

Need a short caption about finding true love?

"True love isn't just about finding someone you can live with—it's about finding someone you can't imagine living without."

What about the trials and tribulations of being the eldest daughter?

"Being the eldest child isn't just about responsibilities—it's also about the "unseen battles" that go unnoticed."

D'you recently join a marathon and need a brief inspirational quote to caption your medal-bearing self-portrait?

"A marathon isn't just about 42 kilometers. It's about proving you can do hard things… and realizing you're capable of more than you ever imagined. 🏁💪"

Or about letting go:

"Forgiveness is not just about letting go of anger or resentment; it's about releasing the power that bitterness has over us."

A haircut technique, even:

"Sectioning isn't just about splitting the hair; it's about setting the foundation for balance, control, and creativity."

All of these are real captions and text overlays I compiled from online.

And while I don't go about exhibiting a crude sense of one-upmanship by shaming people who have these AI-isms in their outputs, it would not not make me internally suspect that their piece, well-intentioned though they may be, is likely the counterfeit creation of a pseudo-thinking machine.

But like I said in the beginning, my aim for writing this is to shift the AI accusation away from em dashes to something else entirely, such as this good ol' trick in persuasive writing.

All while still acknowledging a fundamental fact: that the presence of these so-called AI signs in anyone's content does not 100% prove artificial authorship.

I'd like to mention two runners-up that also ring my AI alarm bells:

The parallelism "From… to…,"

From marketing to sales copies and website blogs to social media captions, this is a favorite expression among LLMs, and I fricking hate it.

Where conversing humans would mostly use it to communicate distance and time range, LLMs tend to use it more often to convey topic or information coverage.

Example: "From daily budgeting apps to sophisticated investment portfolios, managing your money has never been easier."

The conjunctive correlation "Whether… or…,"

In a similar fashion, whether you task Chat to create a thousand words of blog content or a brief summation of something, AI writing assistants abuse this parallel structure to appear balanced and inclusive.

Example: "Whether you're an author, entrepreneur, or business owner, we provide tailored solutions to help you achieve your digital marketing goals."

What about you, dear reader? What is your biggest AI writing pet peeve?