How to spot when writing is AI: The 6 elements of robot style

Updated in June 2026

“Amelia tells me you know when writing is AI or not.”

We were poring over the portfolios and samples of freelance writers in her spreadsheet. I’d already looked through the writing before our meeting, and, well… I was surprised. A lot of the samples showed more than a hint of AI. 

To be clear, I have used (and still use) AI as a collaborator. And perhaps I’m too much of a romantic… but I want the writing to still feel like me. Why choose to be a writer if you eliminate the one thing you got into this for: the actual writing. 

“The signs are subtle,” I told her. “But just as Kafka and Emily Dickinson had a writing style, apparently AI has one too.”

She had a look on her face like ‘this-AI-business-is-giving-me-stress-when-it-should-be-taking-it-away’. “What should I be looking for?” she said.

So I told her what a piece of writing looks like when it’s been ‘assisted’ by ChatGPT, or when the whole brief has been thrown into Claude. 

Here’s what I said to look out for.

1. Excessive use of emdashes

ChatGPT has a staggering number of weekly active users, and is the go-to for many marketers. But ChatGPT has a dead giveaway: a complete infatuation with a punctuation mark. The em dash (—). 

Em dashes add thoughts in the middle of a sentence, and ChatGPT usually squeezes them between phrases—like this—instead of having a space between them — like this.

AI loves to add em dashes to create tension!!and drama!! in its sentences. But 90% of the time they can be cut or replaced by commas.

Of course, writers who love a good em dash now face a horrible catch-22: lose a useful punctuation mark entirely, or potentially be judged for using AI (even if you didn’t).

My opinion? Use them, but question if they’re the most effective choice in each sentence, or whether you’re dropping them lazily where another punctuation mark would work instead.

Look out for:

• "She opened the ancient book—its pages brittle and yellowed—and discovered a map that would change her life forever."

• "Many people believe AI will replace humans completely—but the reality is far more nuanced."

• "Exciting news awaits you this Friday—don’t miss our exclusive announcement!"

2. Forced sass

Is it me, or do AI systems constantly carry Mean Girls energy? When ‘edge-lord’ is dialed up so much I want to vomit, it’s usually a good sign a piece of writing was written by AI.

On LinkedIn it’s particularly bad, because a ‘hot take’ is a common post type where AI can be spotted a mile away. Ta-da phrases like ‘but here’s the truth’ or ‘but here’s what nobody’s saying’ are AI’s way of creating conflict in a piece of writing. Usually, they can be replaced with a simple ‘But’. 

Look out for:

• "But here's the thing:"

• “Then I realized:”

• "The result?"

• "Hot take:"

• “And honestly?”
  

3. AI Buzzwords

A brilliant study was conducted in April 2025 at the University of Helsinki to find changes in student essays from pre- and post-ChatGPT. They found that certain words surged in usage:

Also, student answers grew in sentence length and number of words. 

It’s also alarming how many medical papers began using the word ‘delve’ on PubMed, a clear sign that they’d been at least edited with AI. 

A count of the word “delve” in medical papers. ChatGPT’s first full year was 2023. (credit: Jeremy Nguyen)

In the marketing world, I’ve spotted tons of similar buzzwords repeated in article and content titles: Unlock, Empower, Elevate, or specific title structures: ‘from X to Y’, ‘Master X in X Days’. 

Does this mean that you can no longer use the word ‘crucial’, ‘significant’ or ‘important’, if you really need to? Of course not. But it does mean you should delve into what’s truly crucial, leverage these terms only when needed, and foster a critical connection between the words you use and your message. 😉

Also look out for:

• "Quietly"

• “Grounded”

• "Tapestry"

• "Navigate", or other navigational metaphors.

JRR Tolkien: AI fraud?

4. Cliché phrases

What do 19th-century Romanticism-era poets and AI have in common? 

They are both obsessed with landscapes.

Wanderer above the Sea of Fog - Caspar David Friedrich, 1818

You know a piece of writing is AI when it starts with cliché phrases: “In today’s fast-paced digital landscape’ or ‘In the dynamic world of...’ 

Cliché phrases like this are becoming more commonplace and again, is one of an AI system’s technique to ramp up the emotional drama.

It’s also incredibly excited to talk about ‘looming challenges’ or being incredibly diplomatic much of the time by ‘understanding advantages and disadvantages’.

Look out for:

• “In the dynamic landscape of Cryptocurrency”.

• “As the world continues to evolve…”

• “Shouting into the void”.

• "No fluff".

5. Formulaic sentence structures

AI systems love specific sentence structures, and especially abuse the rule of three. And I get it. Three is symmetry: Sun, Moon, Earth; past, present, future; Father, Son, Holy Spirit…

But AI systems add a third element even when it isn’t necessary. Triplets make sense if all the elements are important and not weirdly broad.

What else should you look out for? Framing statements such as ‘It’s not just X. It’s also Y’ are particularly commonly used by AI systems to show contrast. 

Translation: ‘It’s not just a lipstick, it’s an attitude’.
Spotted in the wild at a Barcelona bus stop. I see you, L'Oréal.

AI also loves breaking things down into structured lists and bullet points, even when it feels too sudden because people apparently hate long text, and that’s the only way we consume information now, right? RIGHT?

Look out for:

• “No [X]. No [Y]. Just [Z].”

• “It is not just [X]. It’s also [Y].”

• Sentences starting with ‘Here’: “Here's why that matters," "Here's the deal."

• “That's the real X.” ie “That’s the real unlock,” “That’s the real challenge.”

6. "As a large-scale language model"

🚩🚩🚩🚩🚩🚩🚩🚩

AI cannot replace a writer’s point of view

In the end, it was impossible to avoid picking a freelancer who we thought wasn’t using AI at all.

And we wouldn’t either.

Because I, and those freelancers, know that progressing in our field (content and copywriting) means learning how to best use AI systems, and learning where they go wrong. Hell, even I use AI to create most of the images on this site (and some writing too!).

By teaming up with AI as a collaborator, I can focus on what I enjoy: researching, exploring the idea, and figuring out how to best communicate it in my own style.

To the naked eye, AI writing is getting harder to detect, and I won’t be surprised if this list is out of date in five minutes. Pretty soon, we’ll all be just like our parents: sharing an article or video that’s AI-generated, and being none the wiser (or simply not caring).

But what we’ll always crave in a piece of writing is an authentic, human voice and point of view. 

That’s irreplaceable, no matter how prevalent AI writing is.