Skip to content
Netanel EliavMarch 7, 2025 9:18:23 AM GMT2 min read

The Typographical Sins of AI: Why Machines Love Em Dashes So Much

The Typographical Sins of AI: Why Machines Love Em Dashes So Much
3:50

 

The Hidden Typography That Reveals AI-Written Content

Have you ever noticed something slightly off about text generated by AI? Something you can't quite put your finger on? It might be lurking in the punctuation—specifically, in those horizontal lines that aren't quite hyphens.

As AI writing tools become increasingly sophisticated, they're leaving behind subtle fingerprints. Among the most telling: an unusual fondness for dashes that most human writers use sparingly.

The Characters Humans Actually Use

Before diving into AI's typographical quirks, let's look at how humans typically punctuate their writing:

Common Top 10 Characters (in order of frequency):

  • Space ( ): ~50%
  • Period (.): ~15%
  • Comma (,): ~10%
  • Question Mark (?): ~5%
  • Exclamation Mark (!): ~4%
  • Quotation Marks (" "): ~4%
  • Apostrophe ('): ~3%
  • Colon (:): ~3%
  • Semicolon (;): ~2%
  • Hyphen (-): ~2%

Notice what's missing? The em dash (—) and en dash (–) don't even make the top ten!

The Dash Dilemma: En Dash vs Em Dash

Let's examine these characters that AI seems to adore:

The En Dash (–) is slightly longer than a hyphen and traditionally used to indicate ranges:

  • "Pages 10–15"
  • "The London–Paris train"

The Em Dash (—) is longer still and typically used to indicate a break in thought:

  • "The meeting-which lasted three hours-accomplished nothing."
  • "She left the party-no one knew why."

Why AI Overuses These Characters

AI systems like ChatGPT, Claude, and others are trained on vast datasets that include academic writing, journalism, and books. These formal sources often use proper typographical conventions that the average person doesn't bother with in everyday writing.

The result? AI produces text with impeccable typographical precision that ironically makes it stand out as non-human.

Consider these examples:

AI-Generated (Typical): "The research-conducted over three years-demonstrated significant results."

Human-Written (More Likely): "The research (conducted over three years) demonstrated significant results." OR "The research, conducted over three years, demonstrated significant results."

The Telltale Signs of AI Writing

Beyond the dash obsession, here are other typographical patterns that often reveal AI-generated content:

  1. Perfect consistency in punctuation usage throughout long texts
  2. Precise spacing after punctuation (always one space, never two)
  3. Correct but uncommon characters like proper ellipses (…) instead of three periods (...)
  4. Formal quotation marks ("curly quotes") instead of straight quotes

Why It Matters for Marketers and Content Creators

As AI detection tools become more sophisticated, these subtle typographical signatures may trigger content filters or undermine authenticity. For marketers trying to create genuine connections, text that subtly reads as "off" can harm credibility.

How to Make AI Writing More Human

If you're using AI tools in your content creation:

  1. Replace some em dashes with parentheses or commas
  2. Vary your punctuation patterns throughout the text
  3. Occasionally use simple hyphens where an en dash would be technically correct
  4. Introduce minor inconsistencies that reflect human writing habits

The Future of AI Writing Fingerprints

As AI systems evolve, they'll likely become better at mimicking human typographical idiosyncrasies. Some are already being trained to vary their style based on casual vs formal contexts.

But for now, keeping an eye on those dashes-particularly those suspiciously perfect em dashes-might help you spot AI-generated content in the wild.

The next time you're reading something that feels just a bit too typographically perfect, look for those dashes. They might be telling you more about the author than the words themselves.

avatar

Netanel Eliav

Netanel Eliav, CTO and Head of AI Research at Jam 7, is a leader in AI, LLMs, and agentic frameworks. From developing security tools as a teen to leading projects at Intel and the Israeli Prime Minister's Office, he drives innovation by merging technology with business strategy. Remember! Knowledge is knowing a tomato is a fruit, wisdom is not putting it in a fruit salad, and intelligence is realising that without tasting it

RELATED ARTICLES