*Updated for ChatGPT 4o, Claude 3.5 Sonnet and Llama 3.1
Here, we look into the word choices AI really likes and better understand the phrases that make your BS detector go off. Once you know what to look for, you'll never unsee it.
How do large language models work?
By now, you probably know all about the likes of ChatGPT and similar platforms. These are large language models, and they’re trained on massive amounts of text data. They learn patterns in language, allowing them to predict likely word sequences.
When you input a prompt, the model doesn't truly understand it. Instead, it generates a response based on statistical probabilities of what words typically follow in similar contexts.
This process can produce human-like text, but it has limitations. LLMs often use overly formal language, struggle with context and can state incorrect information with total confidence even though what it’s telling you couldn’t be further from the truth.
Signs of AI-generated text include:
- Overuse of complex vocabulary
- Awkward transitions
- Lack of original insights
- Repetitive phrasing
- Factual inconsistencies
Some AI writing can be very good, especially when you know how to prompt the specific platform. But when left unchecked, AI can leave a lot to be desired. It sounds overly robotic (because it is), and the content it creates is often bland and uninspiring.
The overachiever vocabulary
Whether it’s ChatGPT, Google Gemini or Meta’s Llama, AI loves to flex its linguistic muscles. It’s probably overcompensating for the fact it isn’t actually a human. If you ever see an overuse of these words in an article, there’s a good chance it’s AI at work.
- "Plethora" - When's the last time you heard a real person say this? AI throws it around like confetti.
- "Myriad" - Same deal. It's the fancy cousin of "lots" that AI can't resist.
- "Utilise" - Because apparently "use" isn't impressive enough.
- "Paradigm" - Eye roll alert. It's the go-to for trying to sound academic.
- "Leverage" - Unless you're talking about actual levers, this is pure business jargon.
The wishy-washy qualifiers
AI has commitment issues. It hedges its bets with vague language:
- "To some extent" - Translation: "I'm not really sure, but I'll say it anyway."
- "In many cases" - Which cases? Who knows!
- "It can be argued" - By who? The AI is covering its bases.
- "Arguably" - See above. It's like a get-out-of-jail-free card for bold statements.
Forced transitions
AI loves to connect ideas, even when it's awkward. Watch out for:
- "Moreover" - The classic "I'm not done lecturing you" transition.
- "Furthermore" - See above, but make it extra fancy.
- "Additionally" - Because apparently "also" isn't good enough.
- "In light of" - Prepare for some pseudo-profound statement.
Overly Formal Phrasing
AI hasn't quite mastered casual speak. It defaults to stuffy language:
- "One might consider" - Who's this mysterious "one"? Just say "you" or "we"!
- "It is worth noting" - As opposed to all those not-worth-noting things?
- "It stands to reason" - Usually followed by something that doesn't actually stand to reason.
- "In the event that" - Just say "if"! We're not drafting legal documents here.
The explainer's toolkit
AI loves to break things down, even when it's not necessary:
- "In other words" - Because the first explanation wasn't convoluted enough.
- "To put it simply" - Ironically, often followed by an not-so-simple explanation.
- "That is to say" - See above. It's the academic version of "like, you know."
- "For instance" - Prepare for an example that may or may not actually clarify anything.
Overeager emphasis
AI hasn't mastered the art of subtlety. It loves to hammer points home:
- "It is important to note" - As opposed to all those unimportant things we've been discussing?
- "Crucially" - Everything is crucial in AI land.
- "Significantly" - The word has lost all meaning at this point.
- "Fundamentally" - Usually followed by something that's not actually fundamental.
The pseudo-expert stance
AI tries to position itself as an authority, often unconvincingly:
- "Studies have shown" - Which studies? By whom? It's the vague appeal to invisible authority.
- "Experts agree" - Again, which experts? It's a sneaky way to add credibility without receipts.
- "It is widely accepted" - By who? The AI hivemind?
- "Research indicates" - See above. It's the academic version of "trust me, bro."
Awkward idiom use
AI struggles with natural-sounding expressions:
- "Hit the nail on the head" - Used way more frequently than any human would.
- "Cut to the chase" - Often used in formal writing where it feels out of place.
- "Barking up the wrong tree" - AI loves animal idioms, apparently.
- "The elephant in the room" - Prepare for an obvious statement presented as profound.
The conclusion crutches
AI hasn't mastered the art of a natural ending:
- "In conclusion" - Because apparently, we couldn't tell it was wrapping up.
- "To sum up" - Usually followed by a summary that's just as long as the main text.
- "All things considered" - What things? We're just repeating ourselves at this point.
- "Ultimately" - The "I'm finally done talking" signal.
Some other words
As well as having its own style of writing, AI does like to overuse single words. If you’ve ever seen a million “ensures” in an article, we reckon AI was at work.
- Delve
- Landscape
- Evolving
- Context
- Insight
- Nuanced
- Perspective
- Paradigm
- Comprehensive
- Supercharge
- Framework
- Facet
- Dynamic
- Intricacies
- Holistic
- Iterative
- Synergy
- Confluence
- Pivotal
- Nuance
- Robust
- Transformative
- Underpinning
- Spectrum
- Trajectory
- In-Depth
- At The Core Of
- A Myriad Of
- On A Broader Scale
- In The Context Of
- From A Holistic Perspective
- Taking Into Account
- A Dynamic Interplay
- Evolving Over Time
- A Comprehensive Overview
- Intricacies Involved
- A Pivotal Role
- Underpinning Principles
- The Spectrum Of
- Transformative Impact
Oh, and “treasure trove”. It really loves using treasure trove for some reason.
Why this all matters
Recognising AI-generated text isn't just about playing "spot the robot." As AI writing tools become more prevalent, it's important to understand their limitations. While AI can be a helpful brainstorming tool or writing assistant, blindly relying on it can lead to stiff, unnatural prose that fails to connect with readers.
For writers and content creators, being aware of these AI tendencies can help you avoid falling into similar traps. It's a reminder to keep your writing grounded, conversational and authentically human.
For readers, recognising AI-generated content can help you critically evaluate the information you're consuming. Is it providing genuine insights, or just regurgitating information in a fancy package?
So all those words are in AI?
Not always. Many of these words and phrases appear in human writing too. What makes AI content stand out is the frequency and context of their use. AI tends to overuse these terms, often in situations where they feel unnatural or unnecessary.
Factors that suggest AI authorship include:
- Consistent overuse across the entire text
- Lack of personal voice or anecdotes
- Odd word choices that don't fit the tone
- Repetitive sentence structures
- Vague or generic statements without specific examples
Human writers might use these words occasionally, but they'll typically vary their language more and provide unique insights based on personal experience.
The future of AI writing
We aren’t here to bash AI, contrary to what you’ve read so far. It's an incredible tool with tons of potential, and we incorporate AI into the work here at Conturae, from strategy to understanding brands and much more. But right now, it's like a really smart toddler trying to sound like a college professor when it comes to relying on it for full-blown writing. It has the vocabulary, but not the nuance or real-world experience to use it naturally.
As AI language models improve, some of these telltale signs might become less obvious. But we’d argue that the goal shouldn't be for AI to perfectly mimic human writing. Instead, we should aim for a future where AI complements human creativity, rather than replacing it.
The most effective writing will always have that human touch—the unique perspectives, emotional resonance and genuine voice that comes from lived experience. AI can help us brainstorm, organise ideas and even polish our prose like it does here at Conturae. But the heart of good writing? That's still a human thing.
Summary: An AI wrote that
The next time you're reading an article that's peppered with "paradigm shifts" and "leveraging synergies," take a moment to consider: are you reading the words of a real person, or just a really impressive language model? And more importantly, does it actually say anything worth reading?
Remember that good writing isn't about impressing people with big words or formal structures. It's connecting, informing and maybe even entertaining. So whether you're writing yourself or working with AI tools, keep it real in every sense of the word.
Want real writing from actual humans who use AI to make their processes smarter? Here at Conturae, we create content that pops for brands so they can strike a chord with their audience.
