Seven Deadly Tells Of AI Writing

Forbes

a day ago

There's a new voice creeping across the internet. It sounds confident, insightful, and polished, made of measured meaningless prose delivered with oh so dramatic flourishes. It's in press releases, social media posts, and student essays. This amateurish easy-to-spot style is characterized by a number of tells which, when revealed, will change the way you read content. I call it The TED-talk style.
The TED-talk style seeks to create engagement, drama, and snappy punch lines that border on insight but are decidedly not insightful. If you know what to listen for, the tells of GPT writing are everywhere. Through limited personal research, I've found many of these same tics in ChatGPT, Perplexity and Grok. In fairness, any writer could have one or more of these ticks, but there are so many to pick from, you can tell.
Here are some examples from a substack post by a well-known marketing influencer with tens of thousands of followers which landed in my inbox Monday. The topic is AI, and it is clearly written by AI. In fairness, only the first ten examples are from his newsletter.
Here are the tells. Starting with the most annoying one.
Contrastive Rhetorical Framing
'So Amazon isn't just buying content. They're buying credibility.'
'This isn't just about revenue diversification. It's about survival.'
'Quality journalism isn't just surviving the AI revolution—it's becoming the premium product that differentiates good AI from bad AI.'
Asking and Answering Rhetorical Questions.
This writing tic has traditionally been associated with bad high-school level essays. Why would one begin a thought with a question? A TED talk needs some drama, but that's contrived in a social media post. For comprehension, context is as important as content. ChatGPT can't tell the difference.
'Financial terms? Undisclosed.'
'What changed? The math did.'
'Why? Because human credibility matters.'
Dashes Everywhere – For No Reason.
ChatGPT favors dashes – over commas. This is the easiest tell – really.
'Amazon can now feed Times articles—plus content from NYT Cooking and The Athletic—directly into Alexa and their AI models.'
'The old model—hire more writers to chase more pageviews to sell more ads—is breaking down in real time.'
Other tics to look out for.
Triplet Framing.
'Fast, cheap, and out of control.' Clever alliteration to a bosa nova beat.
Triples confer rhythm, cleverness and creativity to suggest glib authority. They are often characterized by clever alliteration. It's not necessarily wrong to use triplets but AI uses them compulsively, to wrap up paragraphs with authoritative style.
'Not for advertising. Not for distribution. For AI training.'
The Inspirational Pivot. As easy to spot as contrastive framing. Example: 'This isn't just about AI. It's about humanity.' This is a head fake is to elevate tech talk into TED talk. It shifts from the specific to the abstract to create fake profundity.
Universal Authority Without Source. Example: 'Studies show that storytelling is 22 times more memorable than facts.' Which studies? From where? This tell launders opinion into truth. In this, Perplexity is better than ChatGPT as it returns attributions.
Quotes without attribution. 'AI is the new electricity,' said Musk. Really? I ask the bot. When and where did he say this? When interrogated, ChatGPT admits it is fabricated, and offers to go searching for more quotes. This is the flypaper on which students are often caught. Their critical thinking skills aren't up to challenging the AI.
The reasons writers (and student writers) fall victim to these tropes is that they are human. A student can't recognize the traps: if it sounds good, it is good. Professional writers (not me!) can be lazy, which leads to AI slop. You've written two stories, it's time for lunch, you glance at the prose AI put out and you decide, meh, close enough, even though, on examination, it's not.
Since its launch more than two years ago, ChatGPT's main flaw was delivering overblown, long-winded, and content-free paragraphs. After the March 2025 update, the number of meaningless sentences finally declined, only to be replaced by this very specific TED-talk rhetorical style, which is infecting every model, because they are continuously trained on one another's content and style.
A good TED talk is substantive, fresh, engaging, persuasive and dramatic. Out of context this style is a parody of itself. It is, as author William Faulkner wrote, 'sound and fury, signifying nothing.'