ARC // AI REALITY CHECK
← back to articles
AI Detection · Long-form analysis

You Can't Detect Content Created by AI. You Just Need Attention.

Your tiresome AI-detection flex is a myth. Your posts aren't insight — they're clickbait. And the data proves it.
by C. Pete Connor  •  April 2025  •  ~12 min read

LinkedIn is drowning in a myth: that people can instinctively spot AI-generated content. They can't. But that hasn't stopped hundreds of thousands from posting like they can — each one chasing attention, status, or both. This report cuts through the noise.

I used validated data to examine what's really happening on the platform. As of October 2024, 54% of all long-form LinkedIn posts (100+ words) are likely AI-generated. After ChatGPT launched, AI-authored content spiked 189%. Despite this, a loud minority keeps pushing the same narrative: that their "gut" can sniff out synthetic prose. It's a viral delusion — and it shows up in 12–18% of all AI-related posts, week after week.

Through counterfactual analysis, I explore what the landscape might look like if this myth had never gained traction — or had been publicly debunked at scale. The results point to a sobering conclusion: the claim persists not because it's true, but because it flatters professional egos, feeds algorithmic engagement, and fills the vacuum left by real detection tools that don't work.

This matters. Not just because the myth is wrong — but because it slows down actual solutions, fuels false accusations (especially against non-native writers), and drives content inflation with zero gain in quality. Until the narrative changes, expect the same cycle: performative callouts, junk engagement, and another quarter-million posts pretending intuition is proof.

54%
long-form LinkedIn posts likely AI-generated
+189%
spike in AI content after ChatGPT launch
12–18%
of AI posts repeat the "I can tell" myth
0
detectors that work reliably

Key Takeaway

Between 2023 and 2025, a weirdly persistent belief took hold on LinkedIn: that em dashes (—) are proof a post was written by AI. Not only is that wrong — it's a masterclass in superficial detection theater.

Where It Started (and Why It Spread)

  • Viral bait: Threads called it the "ChatGPT hyphen." Catchy. Baseless.
  • Accessibility fiction: Users claimed humans don't use em dashes because they're "hard to type." Tell that to Emily Dickinson, who used them like oxygen.
  • Status signaling: Spotting em dashes became the new "I'm smarter than the algorithm" flex. It was never about accuracy — just attention.

Why the Claim Falls Apart

  • Historical fact: Em dashes have been around since the 1830s. That's 190 years of human fingerprints on a punctuation mark now being flagged as robotic.
  • Training data truth: AI didn't invent em dashes. It learned them — from us.
  • Stylistic overlap: Humans and AI write similarly because AI was trained on human text. That's not convergence; it's inheritance.
  • Detection failure: If you believe the myth, most 20th-century novels and every op-ed with a breath of style would be "suspicious."

The Bigger Problem

  • Overconfidence bias: People love thinking they're intuitive detectors. They're not.
  • Grammar witch-hunting: Common words and transitions like "furthermore" get tagged as "AI tells." Spoiler: they're just English.
  • Corporate sludge: LinkedIn's content style is so formulaic that humans and bots sound the same. It's not AI's fault — it's the native tone.
"The em dash isn't a sign of AI — it's a sign of a writer who knows how to use punctuation. For me, the em dash isn't just punctuation — it's a lifeline. My brain is always in hyperdrive, with ideas jumping around excitedly and thoughts connecting unpredictably. The em dash allows me to shift gears in my writing without causing a 10-car pileup of confused readers. If the biggest concern people have about authenticity is an em dash, they're missing the point completely. Writing is about connecting — by providing meaning, purpose, and value."
— a writer who actually thinks for a living

Let's stop calling it complicity. LinkedIn didn't just let the human-detector myth spread — it benefits from it, feeds it, and quietly profits off the chaos.

  • House-brand AI with zero accountability: LinkedIn nudges users to rewrite posts with in-app AI tools — then offers no transparency, no tagging, no disclaimer. It's ghostwriting at scale, hidden in plain sight.
  • Conflict equals clicks: The feed rewards tension. "I can tell this is ChatGPT" starts a fight. Fights generate comments. Comments juice the algo. Everyone wins — except the truth.
  • Clout farming as detection theater: Posts that declare "I just knew" get rewarded like thought leadership. The wilder the claim, the better the engagement. It's not accuracy; it's algorithmic cosplay.

Meanwhile, actual solutions — disclosure tools, source metadata, user education — are nowhere to be found. Why? Because solving the problem kills the outrage. And outrage drives traffic.

LinkedIn doesn't just allow misinformation. It's structured to turn it into influence. The platform didn't fail to fix the problem — it built an ecosystem where the problem is the product.

DomainPlausible TrajectoryGrounding in Data
Discourse Tone Fewer "victory-lap" posts about spotting ChatGPT; more neutral conversations on provenance and disclosure. 12–18% of AI-related posts repeat the claim; remove it and a sizable slice of polemic disappears.
Verification Norms Earlier community shift toward objective signals (watermarks, source tags). Current reliance on intuition is traceable to the claim's popularity; absent the trope, users must seek alternatives.
Trust Dynamics Reduced public accusations of "fake authenticity," especially toward ESL writers. Detector bias and false positives well-documented; fewer call-outs lowers collateral damage.
Platform Policy LinkedIn invests in disclosure UX sooner, rather than surfacing unreliable detector scores. Moderation-score visibility changes in Nov 2024 triggered a +189% spike in claim posts; without the claim, that policy feedback loop weakens.
Replacement Narratives Likely rise of "AI-human collaboration" or "quality over origin" frames — still identity-affirming, but less adversarial. Professionals need a status narrative; hashtag clusters serve that role today.

Without the human-detector claim, professional discourse would likely have evolved toward more constructive frameworks centered on transparency and collaboration, rather than detection and conflict.

DomainExpected EffectWhy This Follows
User Behavior "I-can-tell" posts fall well below 5% of AI threads; citation of peer-reviewed benchmarks becomes status currency. Once 650k–975k repetitions lose credibility, repeating it carries reputational risk.
Hashtag Ecosystem #HumanFirstAI and #AuthenticityMatters lose share; tags like #AITransparency gain. Posts under #HumanFirstAI carry the claim 2.3× above baseline. Remove the claim, remove the multiplier.
Algorithmic Ranking Controversy-driven engagement dips; LinkedIn pivots feed incentives toward constructive governance content. Current spikes (+217% Apr 2024, +189% Nov 2024) are engagement gold; debunking cuts that fuel.
Tool Evolution Detector vendors shift from text-only heuristics to provenance metadata; investors follow. High false-positive rates and bias — clear market signals once hype subsides.
Norms Around AI Content Consensus ethic becomes "disclose usage, judge substance." Both automated and human detection are unreliable; disclosure is the logical fallback.

An authoritative debunking would accelerate disclosure mechanisms and provenance solutions, redirecting attention from detection to transparency.

1. It makes people feel smart.

Saying "I can tell" is easier than admitting you have no idea.

2. It gets clicks.

Outrage and fake certainty rack up comments. That's exactly what LinkedIn pushes.

3. It saves LinkedIn from doing real work.

Fake tools, no labels, zero fixes — just vibes.

4. It keeps evolving.

The next lie is, "Well I can tell, even if others can't."

Bottom line: The myth sticks because it feeds egos, farms engagement, and lets platforms dodge responsibility. Nobody cares if it's true — only if it performs.

If you're still flexing that you can "spot AI," congrats — you've noticed what everyone else clocked two years ago and somehow think it's a revelation.

You're not a savant. You're late to the party and yelling about the leftovers like they're breaking news.

The loudest voices claiming AI is killing creativity are the ones spending all their time performing outrage instead of actually creating anything. These are the folks who say, "AI is ruining writing!" — right before they hit post on a LinkedIn monologue made entirely of melodrama, anecdote, and a desperate need for attention. If they spent half as much time honing their craft as they do declaring its death, we might actually get some decent prose out of them.

Let's be real: If 54% of long-form LinkedIn posts are AI-generated, every time you point and guess you've got coin-flip odds. That's not insight. That's roulette in a hoodie.

So no — you didn't "just know." And no — it's not helpful. You're not raising the bar; you're clogging the feed with self-importance dressed as wisdom.

Either get serious about real solutions — transparency, disclosure, actual accountability — or get out of the way.

Because this myth doesn't need more parrots. It needs a funeral.

  1. [1] Half of LinkedIn posts AI-Generated — fudzilla.com
  2. [2] AI Was Born to Blog on LinkedIn — gizmodo.com
  3. [3] Study: Over half of LinkedIn posts now AI-generated — mindmatters.ai
  4. [4] Originality.AI: 54% of long-form LinkedIn posts are AI-generated — originality.ai
  5. [5] Newsbytes: Over 54% of LinkedIn posts are AI-generated — newsbytesapp.com
  6. [6] Do AI Detectors Flag Human-Written Content? — linkedin.com
  7. [7] The Great Hoax: AI-Plagiarism Detectors — linkedin.com
  8. [8] The Truth About AI Detectors — linkedin.com
  9. [9] The Flawed Promise of AI Detection — linkedin.com
  10. [10] OpenAI: AI Text Classifier retired — openai.com
Published April 2025  •  ~12 min read  •  AI Detection · CX · Platform Governance