Key Summary (TLDR)
AI writing “tells” are patterns people’s brains are now on the prowl for as a signal or proof of AI-generated writing.
AI writing tells include things like the em dash (that’s this thingie —), the “it’s not X, it’s Y” structure, the rule of three, and adverbs like “quietly” and “genuinely”.
Here’s the sad but true undercurrent of the AI writing tells. They came from human writing first. And they don’t always, in fact, signify “AI-generated”. Furthermore, scrubbing them out of your work to avoid suspicion can be a losing battle, backfire, or worse… erase the very voice that makes you you!
This article is about:
- The most-flagged AI writing tells right now: em dashes, “it’s not X, it’s Y,” rhetorical questions, the rule of three, staccato “no X, no Y, just Z” lines, adverbs like genuinely and quietly, words like delve and dive, emojis, lists, and a punchy “TED talk” tone.
- AI detectors being unreliable and incorrectly flagging human-written content as “AI”
- How chasing every AI writing tell can flatten your writing into the exact blandness you were trying to avoid. Soon, we’ll all just be writing in beige to avoid appearing like AI. And then, spoiler alert: The AI will just learn that too.This is not a solid nor suggested alternative.
- The real difference between human and AI writing doesn’t come down to a punctuation mark. It’s whether there’s an actual person behind the words which is usually something you can sense and feel.
Prefer to listen? This one started as an episode of Small Business Casual. Listen here.
I never really saw a need for em dashes…
Like a lot of people, I only started noticing them with the rise of AI.
Then, I started noticing them everywhere, and my spidey senses went up, alerting me to one of the first AI writing tells.
These days, I’ve come around. I actually weave them in from time to time (mostly in my personal work and almost never on a website) because they are, in many cases, the perfect little side quest tool. If you’re not familiar with the em dash, it’s a longer version of this “-” and can be used to make an aside to the reader—because lord knows, I might want to take the opportunity to tell you something slightly unrelated in the same sentence every now and then—all in all, a very useful tool.
Sadly, they’re on the chopping block as one of the many punctuation marks signaling AI-generated content.
That brings me to the conversation that kicked this whole research project of mine off.
Someone on a networking call recently shared that she avoids em dashes entirely. If one sneaks in out of habit—or, cough, cough, just decent writing—she has her assistant take it out! A whole punctuation mark is banished from use because she doesn’t want anyone to have even the INKLING that she used AI to write.
I started to wonder…
At what point does avoiding “AI writing tells” stop being good advice and start being a slow surrender of your own voice to a tool you may not even be using?
Humans are Responsible for the “AI” Writing Tells
Spoiler alert, dear human being: All of these so-called AI writing tells came from us! The people! AI learned from human writing. Yup, we invented the em dash, we like to use the rule of three, and we’re the ones guilty of a adverbs, having happily used them for years and years. But now… we’re deleting all this stuff, scrubbing it from our writing, banishing it from our keyboards because a stupid machine (that learned them by copying us) gets the credit when we use these tools.
The Current AI Consensus
There really isn’t one!
Going into this research, I thought it’d be split down the middle. 50% for AI and 50% against. But the latest polls (July 2026 as I write this) tell a more nuanced story.
Get this: Roughly 50% of Americans express more concern than excitement about AI, while only about 16% to 34% are more excited than concerned. About one-third (33%) feel mixed.
I’d put myself somewhere between mixed and highly concerned. Another story for another day.
Why It Matters for Your Business.
When people read, watch, or listen to your content, their brains are subconsciously on the lookout for fakeness and inauthenticity.
Which means… the second something pings as AI, their suspicion creeps in and you potentially begin to erode trust.
What are they hiding? Why couldn’t they be bothered to write this themselves? Why should I keep reading?
Considering that trust IS the point…we don’t exactly wanna’ suck it outta’ the room.
Worth Noting (Before We Roast the AI Writing Tells)
I want to be clear that I’m NOT the AI writing tells police. This isn’t about equipping you to go surveilling your network for signs of robot.
In fact, something to consider is what my friend, Ande Lyons, raised when I posted about this recently on LinkedIn: She reminded everyone to ease up on the judgment. Pre-AI, many people were too anxious or unsure to write publicly. They’re now, thanks to AI, finally able to engage, share their thoughts, and get clarity around their words and message. This also applies to folks with learning differences, neurodivergence, and anyone who went numb staring down a blank page.
It’s an important reminder that our intention is not and should not be “gatekeeping” by slapping them on the hand when we spot or smell AI.
It’s a valid and important point, and that said, this is to raise awareness, not promote accusation. I want you to know the AI writing tells so you can decide what to keep and what to skip, on purpose.
Roll Call: 10 Common AI Writing Tells
The em dash. The poster child of ChatGPT and LLM’s everywhere. A cousin to the parentheses and the ellipsis (the dot dot dots…as my friend Lindsay Dreyer calls them). LLM’s loveeeee them some em dashes, so now humans who have been using them since the dawn of time, get accused of being AI. A little absurd given it’s a perfectly good, centuries-old tool.
“It’s not X, it’s Y.” The sentence structure I see SO SO SO often. It actually is starting to hurt my soul. Example: “It’s not failure, it’s a learning opportunity.” “It’s not you, it’s your strategy.” AI reaches for this constantly.
The rhetorical question. “The worst part?” “The honest truth?” and then off it goes. Especially when the answer involves “something worth sitting with” this sets off my AI writing alarm bells.
The rule of three. Creativity, marketing, and entrepreneurship. Spoons, chopsticks, and sporks. I say I’ve been doing this since I was knee-high to a ladybug. Apparently, now AI is a big fan of it… sigh.
“No X, no Y, no Z. Just this.” No hype, no fluff (AI loves fluff), no noise, just peace. That staccato negative space where a nice idea suddenly reads as a machine.
Adverbs. Ok, this one gets my goat. The greats will tell you to go easy on adverbs anyway and they’re right, butttttt anddddd I love me some adverbs. Truly. I honestly genuinely actually do. Unfortunately for humans everywhere, the ones currently raising red flags are genuinely, honestly, truly, actually, and QUIETLY. I’m saying that last one a little louder, because it’s a big one. A whole part of speech, gone poof… just like that!
Certain vocabulary. Delve. Dive. Straightforward. There’s more. You can learn more in Forbes’ “The 7 Deadly Tells of AI Writing“
Emojis and lists. Apparently formatting is incriminating now too. 🙄 Rest-assured, this blog was written by a human, despite the lists and that emoji. Or, was it!? Dun, dun, dun.
The sass. I’ll show you sass! But seriously, don’t get me started. Why are the robots so sassy? The punchy, dramatic, opinion-riddled tone is taking over the whole internet. Forbes calls it a “TED Talk voice”. All punch lines, built for drama and mic drops.
“They’re not just doing X, they’re doing Y.” “Amazon isn’t just buying content. They’re buying credibility.” “This isn’t just about revenue diversification. It’s about survival.” You’ve seen it a hundred times today, I’m sure.
NEWSFLASH: Half of that list is just good, human writing. Rhythm? Contrast? A well-placed aside? Yeah, we wrote that first! We didn’t install those techniques into the robots, and they most certainly didn’t invent adverbs.
So What Are We Supposed to Do!? Delete Everything?
Personally, I’m considering suing for emotional damage. But, after that… I don’t plan to rebel entirely and riddle my text with typos or fill my sentences with adverbs, I plan to keep calm and carry on (mostly).
There’s an Atlantic piece by Adam Kirsch called “Why AI Is Incorrigibly Didactic” that compares our tell-hunting to a long-discredited bit of 19th-century pseudoscience.
You see, back then, people believed you could spot a criminal by the shape of his jaw and cheekbones. Adam’s point is that treating the em dash or the “it’s not X, it’s Y” as proof of a machine is the same kind of pseudoscience as judging what’s inside by what’s outside.
Using the AI writing tells as actual evidence of AI doesn’t hold up, because a warm-blooded, true blue, living and breathing human can STILL love an em dash.
It’s a style choice, not a confession. (see what I did there?)
As the models continue to improve, they’ll only get better at mimicking us anyway.
I’ve even heard people start adding deliberate typos to their content to ENSURE people know it’s human which is frankly, a losing battle… the machines will simply learn to copy that, and any other tricks we’ve got up our sleeves, too.
These patterns are a symptom, not the disease.
Cut the em dash, then delve, then adverbs, then the rule of three, then emojis, then lists. Obey every rule, and before you know it, you’ve sanded off everything that sounded like you!
The more everyone scrubs the same tells, the more uniform all human writing gets. We’re editing toward identical blandness to prove we aren’t bland.
Even the AI Detectors Can’t Spot the Tells Well…
I ran some of my own pre-AI-era writing through an AI checker. It came back anywhere from 11-46% AI-written. Yup, that’s on pieces I wrote every single word of before AI was wearing diapers.
So even the damn AI can’t reliably tell what’s AI! Which makes me wonder whether the writing tells are worth all this fuss in the first place.
What Actually Matters
Here’s what I think is most important for business owners to remember when it comes to AI.
- AI brain rot is real. The more you outsource your ideas and writing to AI, the more you erode your creative ability and thinking for yourself. It’s becoming a real problem and something I don’t think we should take lightly! As my fellow podcaster Joe Casabona puts it, “I’m using computers for computer things. I’m not letting it make the final decision.” He’s got a great piece called “3 Signs You’ve Got AI Brain Rot (and How to Prevent It)” that’s well-worth the read.
- 2. The snake eating its own tail problem of AI. As more AI content floods the internet, AI starts training on and regurgitating its own output, which just multiplies more and more of the same. Forget turtles… It’s sameness all the way down. (far less cute than turtles)
- Protect your heart, soul, personality, and you-ness with everything you’ve got! Ask yourself why you’re outsourcing your voice in the first place. Just trying to save time? If it’s costing you clients and trust, is that a good trade? Stuck creatively? What did you do to get unstuck before AI existed?
AI Writing Tells are the Surface…
For me, the AI writing tells are surface-level. Semantics.
I believe that the difference between humans and AI can’t be distilled down to a punctuation mark and a sentence structure. The deeper issue here is the heart and soul, and I still believe we can detect rather well whether or not there’s a real person behind the words. We can sense when it’s a real-life perspective we’re getting or one of Claude’s outrageous analogies.
If you take nothing else away from this article, I invite you to consider this:
You don’t prove you’re human by deleting em dashes and Oxford commas. You prove it by being the version of you that can’t be faked. How do you become more of THAT version of you?
Instead of asking, “How do I avoid sounding like AI?” Start asking, “How do I sound unmistakably, adverbly, quirkily, absurdly, unabashedly like me?“
That question will take you places the first one can’t.
FAQ: AI Writing Tells
In case you’ve still got questions after all that. Or you just scrolled down here wondering if the good stuff was at the bottom.
What are AI writing tells?
+ AI tells are patterns readers now associate with AI-generated text: the em dash, the “it’s not X, it’s Y” construction, rhetorical questions, the rule of three, staccato “no X, no Y, just Z” phrasing, adverbs like genuinely, honestly, and quietly, words like delve and dive, and an overly punchy, sassy tone. The list goes on…
Do em dashes mean something was written by AI?
+ Ahem, no, no, no, no, dear goodness, no. The em dash is a normal, highly useful punctuation mark that writers have used for centuries. When ChatGPT first came onto the scene, it was used a lot, and it quickly became associated with AI. But ask 70% of human writers, and they’ll tell you it’s not an indication of a robot but a beloved style choice.
Are AI detectors and checkers accurate?
+ Not reliably. They often produce false positives, flagging entirely human writing as partly AI-generated. In one test, my own 100% human work (pre-AI-era) came back as 11-46% AI. Treat detector scores with heavy skepticism, and don’t let one edit your voice.
Should I stop using these things so my writing doesn’t look like AI?
+ Only if that perception matters for your goals, and even then, I’d encourage you not to overcorrect. Scrubbing every so-called AI writing tell tends to flatten your writing into the exact generic blandness you were trying to avoid. A better focus is sounding unmistakably like yourself, with specific stories and details only you could write. That’s the thing AI can’t fake.
