Table of Contents show

If there is one punctuation mark that has defined the “AI Era” of academic writing, it certainly not the period, and it certainly isn’t the Oxford comma. It is the long, interruption-heavy line known as the em dash (—).

In 2026, if I open a thesis from a student at UM or USM and I see three em dashes in the first abstract paragraph, I don’t even need to run it through Turnitin. I already know. You prompted a Large Language Model (LLM), you copied the output, and you pasted it directly into your manuscript.

It has become the “vocal fry” of digital writing — an affectation that sounds sophisticated to the machine but exhausting to the human examiner.

To the AI, the em dash is a useful connector. To a Malaysian supervisor, it is a red flag. It screams, “I didn’t write this; I generated it using AI.”

In this post, we are going to dissect the Em Dash. We will also explore why algorithms are addicted to it, why it ruins your academic flow, and how you can perform a “Punctuation Detox” to save your AI score.

1. The Anatomy of the Mark: What is an Em Dash in the first place?

Before we analyze why the bots love it, we have to understand what it actually is.

In the hierarchy of horizontal lines, the em dash is the biggest brother of them all.

The Hyphen (-): The shortest line. Used for compound words (e.g., state-of-the-art, well-known). It joins things.
The En Dash (–): The middle child. Roughly the width of the letter “N.” Used for ranges (e.g., 1990–2026, pp. 45–50). It implies “through” or “to.”
The Em Dash (—): The longest line. Roughly the width of the letter “M.”

The Function of “Dramatic Interruption”

In traditional English grammar, the em dash is the “Wild Card.” It is incredibly versatile. It can do the job of a comma, a colon, or a pair of parentheses — but with a more dramatic effect.

Comma version: “The results, which were unexpected, changed everything.” (Soft, flowy).
Parentheses version: “The results (which were unexpected) changed everything.” (Quiet, whispered).
Em Dash version: “The results — which were unexpected — changed everything.” (Loud, emphatic).

The em dash creates a hard break in the sentence. It forces the reader to stop, take a breath, and look at the information sandwiched between the lines. It is a tool for emphasis.

And that is exactly why the AI loves it. It thinks everything deserves emphasis.

2. The “Sophistication Trap”: Why LLMs Are Addicted to the Dash

If you ask ChatGPT, Gemini, Claude, Deepseek or Perplexity to “rewrite this paragraph to sound more academic,” the first thing they will do is inject an em dash. Why?

It comes down to how Large Language Models (LLMs) are trained and how they predict text. Because the data sets they were trained on contain loads of them!

The “New York Times” Bias

LLMs are trained on vast datasets of “high-quality” text from the internet. This includes millions of articles from outlets like The New York Times, The Atlantic, The New Yorker, Wikipedia and high-level academic journals.

In high-end journalism and creative non-fiction, the em dash is a stylistic darling. Writers use it to sound thoughtful. Because the AI is trying to mimic “smart writing,” it statistically associates the em dash with “high intelligence.”

When you ask the AI to write a thesis introduction, it defaults to this “Intellectual Persona.” It thinks: “Smart people use dashes to insert complex thoughts; therefore, I will use dashes to show I am smart.”

The “Safe Connector” Theory

This is the technical reason. LLMs work by predicting the next word (token) in a sequence.

When an AI is generating a long, complex sentence, it often needs a way to tack on an explanation without starting a new sentence (which requires planning a new grammatical structure) or using a clunky transition word (like “however” or “furthermore”).

The em dash is the ultimate “Safe Connector.”

It doesn’t require a conjunction (like “and” or “but”).
It doesn’t require specific grammar rules (like a semicolon does).
It allows the AI to simply append more information.

Example of AI Logic:

Generator: “The qualitative data suggested a correlation…” AI Decision: I want to add more detail about the data source. Option 1 (New Sentence): “. This data came from rural clinics.” (Too choppy). Option 2 (Comma): “, which came from rural clinics,” (A bit weak). Option 3 (Em Dash): “— gathered primarily from rural clinics —” (Perfect. It connects the thought instantly and keeps the sentence moving).

To a probability engine, the em dash is the path of least resistance. It is the “duct tape” of AI writing — it holds complex sentences together without needing complex grammar.

3. The Supervisor’s Radar: Why It Looks Like “Bot-Speak”

So, why is this a problem? If the New York Times uses it, why can’t a PhD student at UKM use it?

The problem is Frequency and Context.

The “Breathless” Effect

Human writers use em dashes like salt — a pinch here and there to enhance the flavor. AI uses em dashes like rice — it is the main bulk of the meal.

When a supervisor reads a thesis where every paragraph contains a sentence structure like:

“Subject — elaboration — verb.”

…it creates a breathless, jerky reading experience. It feels “manufactured.” Humans naturally vary their sentence structure. We use commas. We use periods. We use brackets. When we see a page with 15 em dashes, it looks visually distinct. It looks like a pattern.

The “False Nuance”

The most annoying habit of AI is using the em dash to “mansplain” a concept that doesn’t need explaining.

The AI Sentence:

“The study utilized a quantitative approach — specifically a survey method — to gather data.”

The Human Reaction: We didn’t need the dash there. “Specifically a survey method” isn’t a dramatic revelation; it’s just a detail. A human would write: “The study utilized a quantitative survey method.”

The AI uses the dash to create False Nuance. It tries to make a simple sentence look complex by breaking it apart. To a seasoned Malaysian examiner, this looks like “padding.” It looks like a student who is trying to inflate their word count or sound smarter than they are.

The Formatting Glitch

Finally, there is a dead giveaway that happens often in Malaysian universities.

Different word processors handle dashes in different ways.

Word might auto-correct -- to —.
Google Docs might keep it as --.
LaTeX might turn it into ---.

When a student copies from ChatGPT (which renders dashes perfectly) and pastes it into a Word doc that has different formatting settings, the dashes often appear in a slightly different font or size than the surrounding text. Or, worse, the student mixes them up.

I have seen theses that contain:

Hyphens (-) for compound words.
En dashes (–) for page numbers.
Em dashes (—) for AI-generated sentences.
Double hyphens (–) for sections the student wrote themselves.

This “Frankenstein Punctuation” is immediate proof of copy-pasting. It tells the examiner you didn’t just use AI; you didn’t even care enough to proofread the formatting.

4. The “Mid-Sentence Pivot”: A Case Study

Let’s look at a specific example of how this destroys academic flow.

The Prompt: “Write a problem statement about dengue fever in Selangor.”

The Likely AI Output:

“The rising incidence of Dengue in Selangor — a state known for its rapid urbanization and high population density — presents a critical challenge for public health officials — who are struggling to contain the vectors.”

The Critique: Look at that sentence. It is a “Russian Doll” of interruptions.

It starts with a subject (Dengue).
It interrupts itself to define Selangor (— a state known for… —).
It returns to the main point (presents a challenge).
It interrupts itself again to explain the officials (— who are struggling…).

This is Linear Thinking. The AI is generating the thought in real-time, adding qualifiers as it goes.

The Human Rewrite (Scholarly):

“The rising incidence of Dengue in Selangor presents a critical challenge for public health officials. As a state characterized by rapid urbanization and high population density, Selangor faces unique difficulties in vector containment.”

Do you feel the difference? The human version groups the ideas logically. It uses a period to separate the “Problem” from the “Context.” It is authoritative. The AI version is a run-on stream of consciousness held together by dashes.

If you are currently staring at a draft that is littered with long, horizontal lines, do not panic. You don’t need to delete them all. You just need to translate them back into human academic English.

Here is your “Punctuation Detox” protocol.

5. The “Translation Protocol”: How to Fix It

To de-bot your thesis, you need to replace the em dash with the punctuation mark the AI was trying to avoid.

The “Mansplaining” Dash

The Bot: “Qualitative research — which focuses on non-numerical data — was chosen for this study.” The Fix: Use commas. The Human: “Qualitative research, which focuses on non-numerical data, was chosen for this study.” Why: Commas are standard. They don’t draw attention to themselves. They let the sentence flow.

The “Afterthought” Dash

The Bot: “The results were significant — p < 0.05.” The Fix: Use a semicolon or parentheses. The Human: “The results were significant (p < 0.05).” Why: In academic writing, statistical values usually belong in parentheses, not tacked onto the end of a sentence like a dramatic reveal.

The “List” Dash

The Bot: “We analyzed three factors — cost, time, and quality.” The Fix: Use a colon. The Human: “We analyzed three factors: cost, time, and quality.” Why: The colon is the correct grammatical tool for introducing a list. The AI avoids colons because they require a complete independent clause before them. Humans know how to use them correctly.

6. The “Ctrl+F” Audit: Your Pre-Viva Checklist

Before you submit your thesis to IPS (Institute of Postgraduate Studies), perform this 2-minute audit. Of course if you used AI content, you would need to humanize your work as well. It will save you hours of explaining later.

Search for “—” (Em Dash): If you find more than 3 per page, you have a problem.
Search for ” – ” (Space-Hyphen-Space): This is the “lazy” em dash. If you see this, it means you didn’t even bother to format the AI output. Change it immediately.
Check the Font: Zoom in to 200%. Is the dash in Segoe UI while your text is in Times New Roman? If yes, select the whole paragraph and “Clear Formatting.”

7. When to Actually Use It (The “Human” Way)

I am not saying you should never use an em dash. I am saying you should earn it.

In 2026 academic writing, the em dash is a “Power Move.” It is used to emphasize a disruptive or contradictory finding.

Good Usage:

“The literature predicted a linear growth model. Our data suggested the opposite — a complete collapse of the variable after Week 4.”

Here, the dash works. It creates a pause. It highlights the shock of the finding. It isn’t just connecting two boring clauses; it is spotlighting a discovery.

The Golden Rule: Use an em dash only when you want the reader to physically stop and look at what comes next. If the information is just a detail, use a comma.

Conclusion: Punctuation is Your Fingerprint

In your academic writing, your punctuation is your own unique fingerprint.

An AI writes in a flat, predictable rhythm, glued together by endless em dashes. A human writes with variety. We use short sentences. We use colons. We use semicolons (sparingly). We treat punctuation not just as a way to connect words, but as a way to control the breath of the reader. AI has low burstiness, or sentence length variation. It is monotonous. Humans, on the other hand, write with high burstiness. They vary sentence length naturally.

Don’t let a chatbot dictate the rhythm of your research. Reclaim your comma. Master the colon. And leave the em dash for those rare moments when you truly have something dramatic to say.

Humanizing AI-Generated Content: How to Achieve a Low AI Score on Turnitin AI Checker and Why It Matters for Postgraduate Students

The Inevitable Em Dash — The “Giveaway” of AI Writing