17 Dead Giveaways That AI Wrote Your Content (And How to Fix Them)

And no, the dash isn’t one of them.

We’ve reached an odd point in writing culture. Everyone has access to the most powerful text generator ever made, and yet the web is filling up with words that sound like they came from the same mind. That sameness isn’t coincidence, though, it’s data‑driven.

Stop sounding like everyone else using ChatGPT

I’ve been writing for many years, and with an LLM beside me for the better part of 2025. From experience, the technology has been incrementally better on a nearly weekly basis — so it’s counter-intuitive to me that it’s becoming easier and easier to spot weak-effort AI-generated text.

Let me be clear — I don’t believe that AI assistance is wrong. If you’ve got something to say or write, you should be able to use a tool, if you need to, to express yourself. But everyone having access to the same tool and using it in the same way makes that tool and the effort essentially worthless.

Research from Penn State found that humans can distinguish AI-generated text only about 53% of the time — barely better than random guessing. Even with training, accuracy doesn’t improve much. Yet people confidently claim they can spot it every time.

Problematic Origins in the Training Data

AI writing tools are powerful, but they have a default voice that’s formal, balanced, clinical and ’safe’. They default to:

• A formulaic structure
• Using’ motivational poster’ language
• Over-explanation
• Predictable transitions
• Perfectly balanced coverage

Most AI writing models are trained on vast corpora of blog posts, LinkedIn articles, Substack essays, and tech journalism — all genres that heavily use those phrases to signal depth, curiosity, or opinion. Over time, these expressions become statistically high-probability patterns that language models reproduce whenever a human-like or “thought-leadership” tone is requested.

Here are 17 patterns I’ve identified (I’m certain that more exist) that give away AI writing, why they happen, and how to eliminate them — so you can use AI without sounding like everyone else’s content-marketing bot.

— -

1. Dramatic Reveal Phrases

The Pattern:
AI loves to create artificial suspense that promises ‘insight’ but then delivers nothing. It’s hype. I say “Bleah” to hype for hype’s sake.

Common tells:

• “The answer surprised me:”
• “Here’s what blew my mind:”
• “The crazy part?”
• “Plot twist:”
• “Delve into” (research shows this phrase increased 25-fold in 2024)
• “Navigate the landscape”
• “Embark on a journey”

Why AI uses these:
LLM training data includes clickbait and engagement-optimized content, so AI thinks that this is compelling writing. Is it? “HERE’S THE TRUTH ABOUT DRAMATIC REVEAL PHRASES THAT OPENAI DOESN’T WANT YOU TO KNOW”

The fix:
Just state your finding directly. That’s it.

❌ “The answer surprised me: clarity mattered more than content.”
✅ “After all that, I found out that the clarity of my message mattered more than the content.”

— -

2. Formulaic Comparisons

The Pattern:
Constant “It’s not X, it’s Y” structures.

Common tells:

• “It’s not X, it’s Y”
• “The real [noun] isn’t X — it’s Y”
• “Don’t do X. Do Y instead.”

Why AI uses these:
This structure dominates thought leadership content. It’s simple contrast, easy to replicate.

The fix:
Use once if genuinely useful, then vary your structure.

❌ “The value isn’t task completion, it’s priority identification.”

❌ “The goal isn’t output generation, it’s framework creation.”

✅ “Taking time to identify priorities wins out over task completion. I’m building frameworks, not just generating outputs.”

— -

3. Motivational Poster Language

The Pattern:
Sweeping declarations about “winners” and “the future.”

Common tells:

• “This is what separates X from Y”
• “The winners will be those who…”
• “This changes everything”
• “The future belongs to…”

Why AI uses these:
Business books and LinkedIn posts are full of this language. AI thinks it sounds authoritative.

The fix:
Make specific claims. Drop grand pronouncements.

❌ “This is what separates people who use AI from people who master it.”
✅ “That’s the difference between people who merely use these tools and those who are mastering them.”

Even better: Just make your point already — without the comparison.

— -

4. Emoji Headers Everywhere

The Pattern:
Organizing everything with emoji-prefixed headers.

Common tells:

• 🎯 Goal
• 💡 Key Insight
• ✅ Action Item
• 🚀 Next Steps

Why AI uses these:
Modern content marketing uses this style. AI sees it as good structure. Emojis can broaden accessibility for users who rely on visual or emotional cues, supporting comprehension in casual digital content.

The fix:
Use actual headers and write in paragraphs when appropriate. Emojis can actually hinder accessibility for users with screen readers, low vision, or cognitive processing differences if misused.

❌ “🎯 Goal: Build better systems
💡 Insight: Context matters
✅ Action: Document decisions”

✅ “Over time you will build better systems by documenting your decisions. Context matters.”

The most inclusive approach is to treat emojis as emotional punctuation — never as semantic content — and to position them in a way that maintains clear, linear readability for all audiences.

— -

5. Meta-Commentary Placeholders

The Pattern:
Obvious placeholders revealing template-following.

Common tells:

• [deep dive]
• [rabbit hole]
• [hot take]
• [call to action]
• [insert example here]

Why AI uses these:
Phrases like “deep dive,” “rabbit hole,” and “hot take” have become common across AI-generated writing because they mirror recurring linguistic fingerprints in the datasets used to train LLMs and reflect the performative tone of online discourse

These idioms are also part of the post-2010 social-media vernacular — derived from podcasting, YouTube commentary, and Reddit culture. They convey casual authority, giving machine‑generated prose a performative intellectualism that mimics creators’ voices. In essence, the model isn’t trying to be cliché; it’s reproducing the linguistic norms of the human internet that trained it. It is, in essence, parroting an archaic slice of our culture, with a limited vocabulary.

The fix:
Never publish with these in. Just don’t. Write actual content or remove the section.

— -

6. Manufactured Enthusiasm

The Pattern:
Intensity modifiers and dramatic questions that don’t realistically — or at least shouldn’t — match how humans express genuine excitement.

Common tells:

• “And honestly?”
• “The real game changer?”
• “You absolutely need to…”
• “This is incredibly important”
• “I was blown away by…”
• Excessive italics for emphasis
• “And for the first time…”
• “That’s when it clicked.”
• “Fast forward to today:”

Why AI uses these:
Emphasis markers and rhetorical questions appear frequently in persuasive writing and personal narratives. AI overuses them because it can’t distinguish genuine excitement from generic intensification.

The fix:
Use emphasis sparingly. Real excitement comes through word choice and specificity, not formatting or rhetorical questions.

❌ “And honestly? That’s when it clicked. This was incredibly important.”
✅ “I realized the approach needed to change.”

✅ “I ran screaming naked through the streets, proclaiming ‘Eureka!’ until they caught me and took away my sugar.”

Okay, that last one’s a ‘maybe’.

— -

7. Over-Structured Lists

The Pattern:
Everything becomes numbered lists, even when prose works better.

Common tells:

• Lists within lists within lists
• (Within lists)
• “Here are the 7 ways to…”
• “3 key takeaways:”
• Every paragraph starts with a bullet

Why AI uses these:
Lists are easy to generate and look organized. Plus it’s a lower risk than flowing prose — lists are data, and LLMs can output that better than paragraphs at the end of the day. But there’s also a generation of readers trained on click-bait listicles, who will still click out of instinct.

The fix:
Write paragraphs for flowing arguments. Save lists for discrete items that don’t build on each other. Or, if I can put it another way — write as you would like to be read to.

— -

8. Generic Transition Phrases

The Pattern:
Same transitions repeatedly, creating robotic rhythm.

Common tells:

• “Here’s the thing:”
• “Let’s be clear:”
• “The bottom line:”
• “At the end of the day:”
• “The reality is:”

Why AI uses these:
These phrases dominate conversational business writing. AI thinks they feel informal and direct.

The fix:
You need to vary your transitions, heck, sometimes you don’t even need one — you just need to make your next point and move on.

❌ “Here’s the thing: AI can’t replace judgment. The truth is: every decision was mine.”
✅ “AI just cannot replace actual, opinionated judgment. Every decision was mine, and mine alone.”

— -

9. Artificial Narrative Arc

The Pattern:
Forcing story structure into non-narrative content.

Common tells:

• “So [thing happened] and honestly?”
• “I wasn’t expecting this, but…”
• “At first I thought X, but then…”
• “The journey taught me…”

Why AI uses these:
Personal essays have these elements, but AI applies them everywhere, even to analysis — and honestly? (I’m kidding, I’m kidding)

The fix:
If you’re writing analysis, present analysis. If you’re writing a story, save the narrative for that.

— -

10. Constant Hedging

The Pattern:
AI hedges everything to avoid being wrong.

Common tells:

• “It’s worth noting that…”
• “You might find that…”
• “In some cases…”
• “Generally speaking…”

Why AI uses these:
AI doesn’t want to make incorrect claims, so it softens everything for safety’s sake.

The fix:
Make direct claims when you can support them. Hell, most of the LLMs now have citable research built into them. Ask them for help — do your research, and form an opinion.

❌ “It’s worth noting that, in many cases, you might find that AI generally works better when…”
✅ “AI works better when…”

— -

11. Template Markers

The Pattern:
Phrases revealing template-following.

Common tells:

• “In this post, we’ll cover…”
• “By the end of this article, you’ll…”
• “Let’s dive in”
• “Without further ado”

Why AI uses these:
Blog templates and SEO content use these extensively. AI thinks they’re required, and in many cases when writing for SEO/GEO content, they still are — but not for everything.

The fix:
Just start. Your reader can see the structure.

— -

12. False Specificity

The Pattern:
Numbers and details that sound specific but aren’t.

Common tells:

• “Studies show…” (no citation)
• “Research indicates…” (no source)
• “Experts agree…” (which experts?)
• “X% of people…” (made-up statistic)

Why AI uses these:
They sound authoritative. But when AI lacks actual data, it fakes specificity. It’s hard to believe that this is still an issue today, but it persists. I’ve actually been on the receiving end of a citation and link for a paper that had its publishing date set 6 months into the future.

The fix:
Provide real citations or acknowledge limitations. “In my experience” is more honest than “studies show” when you don’t have studies. As I mentioned earlier, your LLM can probably do research — and if not, well, Google’s still a thing, right? It’s never been easier to find data and facts to help get specific. If you see stuff like this, the author is sorely lacking credibility and either they haven’t noticed — or they’re hoping that you won’t.

— -

13. Overly Formal Clinical Language

The Pattern:
Formal clinical phrasing when plain language works better.

Common tells:

• “Individuals with diabetes” vs. “people with diabetes”
• “Glucose” vs. “sugar” (AI uses this at 2x human rate)
• “Utilize” vs. “use”
• “In order to” vs. “to”

Why AI uses this:
Training data includes academic papers and clinical research. AI thinks formality equals professionalism.

The fix:
Use plain language. If you wouldn’t say it in conversation, don’t write it.

❌ “Individuals seeking to optimize glucose regulation should implement systematic dietary interventions.”
✅ “People with diabetes should seek professional guidance on their diet.”

— -

14. Opening and Closing Clichés

The Pattern:
Specific phrases for opening and closing sections.

Common tells:
Openings: “Moreover…”, “Furthermore…”, “In today’s digital world…”
Closings: “At the end of the day…”, “In conclusion…”, “Moving forward…”

Why AI uses these:
Common academic and business transitions. AI thinks they create structure.

The fix:
Delete them. Your point is clearer without mechanical transitions.

❌ “Furthermore, it’s important to note that, at the end of the day, systematic approaches matter.”
✅ “Systematic approaches matter, as we’ve seen in the arguments.”

— -

15. The Balance Problem

The Pattern:
Every point gets equal weight, which creates unnaturally even paragraphs.

What this looks like:

• Every section is roughly the same length
• All topics are covered comprehensively, even minor ones
• There’s no natural emphasis or de-emphasis
• Missing “this is the important part” signal

Why AI does this:
AI doesn’t have opinions or priorities out-of-the-box, so it gives balanced coverage because it can’t tell what matters more.

The fix:
Real writing has a rhythm. Some points might deserve three paragraphs. Some might deserve more but get edited down for the sake of brevity. Some deserve one sentence. The structure of your writing will reflect importance — not everything needs to be perfect!

— -

16. Paired Adjective Obsession

The Pattern:
Unnecessarily pairing adjectives.

Common tells:

• “Unique and intense”
• “Comprehensive and thorough”
• “Simple and straightforward”
• “Complex and nuanced”

Why AI does this:
When paired adjectives appear in descriptive writing, it’s because AI thinks “more adjectives” equals “more descriptive”.

The fix:
Pick one — usually the stronger one.

❌ “The approach is comprehensive and thorough.”
✅ “It’s a thorough approach.”

— -

17. Relentless Sentence Uniformity

The Pattern:
Every sentence is short. Every sentence is punchy. Every sentence makes its point efficiently. It’s exhausting. See how exhausting this gets?

Common tells:

• Consistent sentence length (usually 10–15 words)
• No run-on sentences ever
• No sentences that breathe and wander
• Everything optimized for “clarity” and “scannability”
• Missing the natural rhythm of human thought

Why AI does this:

This just follows content marketing guides and Hemingway App worship. AI thinks using short sentences is good writing, and it confuses clarity with monotony.

The fix:
Vary your sentence length dramatically. Sometimes you need a short punch. Sometimes you need a sentence that meanders and breathes and takes its sweet time getting to the point because that’s how human brains actually work when they’re thinking through something complex or trying to capture a feeling that doesn’t fit neatly into a tidy little package.

❌ “AI writes simply. It uses short sentences. This makes things clear. But it gets boring fast.”

✅ “AI writes in these relentlessly uniform short sentences that technically make things clear but somehow manage to be both exhausting and boring at the same time — which is impressive in its own terrible way.”

The truth:
Good writing has rhythm; jazz has rests. So does prose.

— -

The Self-Check Process

Before publishing AI-assisted content, run this audit:

1. Search for red-flag phrases:

• “the answer surprised me”
• “here’s the thing”
• “it’s not X, it’s Y”
• “this is what separates”
• “and honestly?”
• “that’s when it clicked”

2. Check patterns:

• Same transitions repeatedly?
• Lists within lists?
• Emoji headers everywhere?
• Every section starts with “Not this/But this”?
• All sentences roughly the same length?

3. Read it aloud:

• Does it sound like you talking?
• Or like a motivational poster?
• Would you say these sentences in conversation?
• Does it have rhythm, or does it march in lockstep?

4. Compare to your previous work:

• Look at pre-AI writing
• Does AI-assisted work match that voice?
• If not, edit until it does

— -

The Real Solution

The best way to avoid LLM tells isn’t memorizing this list. It’s:

1. Have a strong existing voice — AI should be enhancing it for you, not re-defining it
2. Edit ruthlessly — AI helps with first drafts, you own the final version. That’s right, you still need to put the time and effort in
3. Read widely — Know what good writing sounds like in your domain
4. Trust your ear — If it sounds weird, it probably is
5. Embrace imperfection — Humans make weird word choices. We use run-on sentences, we throw in the occasional non-sequitur. That’s not a bug, that’s what makes us unique.

Yes — use AI to generate ideas. Use it to structure arguments, and catch mistakes. But keep your voice — or take the time to define it — including the parts that aren’t optimized for “clarity” or “scannability.”

Perfect prose is suspicious. Give yourself permission to be messy. Bleep, blorp.

— -

Would You Like to Know More, Fellow Human?

I publish Signal Over Noise every week — filtering what actually matters in AI from all the hype and buzzword garbage. No revenue theater, no bullshit metrics, just an honest analysis of what works and what doesn’t.

Subscribe: signalovernoise.at