Why AI Loves Em Dashes—And How to Stop It

You've noticed it. That telltale punctuation mark that turns every AI response into a wall of dashes.

"This is an important point—and one worth exploring—because it reveals something deeper about how these models work—which is worth understanding."

Em dashes are everywhere in AI-generated text. ChatGPT uses them. Claude uses them. Gemini, Grok, Copilot—all of them. And once you see it, you can't unsee it.

The good news: ActMoreHuman fixes this with a single click.

What's an Em Dash, Exactly?

The em dash (—) is a punctuation mark used to signal a pause, an aside, or an abrupt change in thought. Used sparingly, it's a legitimate stylistic choice. Used the way AI uses it—constantly, in every sentence, for everything—it becomes a dead giveaway that a machine wrote your content.

Real writers use em dashes maybe once or twice per page. AI models use them multiple times per paragraph.

Why Does AI Overuse Em Dashes?

AI language models learn writing patterns from massive amounts of text scraped from the internet. A lot of that text comes from sources—Wikipedia, academic writing, journalism—that use em dashes more frequently than everyday conversation.

When these models are trained to sound "formal but readable," they land on the em dash as a shortcut. It signals sophistication. It creates visual rhythm. The problem is that they apply it everywhere, without the restraint a human writer would have.

The result is prose that feels mechanical precisely because of the punctuation that was supposed to make it feel polished.

The Real Cost of AI Em-Dash Overuse

Here's what em-dash saturation does to your writing:

It signals AI authorship immediately. Editors, clients, and readers have all been trained to spot this pattern. One em-dash-heavy paragraph and the game is up. It disrupts reading flow. Real em dashes create emphasis through contrast—they work because they're unexpected. When every sentence has one, none of them do. It undermines your voice. If you write casually, professionally, or with a specific cadence, AI's em-dash habit overwrites that with a generic "AI voice."

Before and After: Em Dash Overload vs. Actual Human Writing

Before (Default AI Output)

"Remote work has changed everything—from where we work to how we communicate—and companies that haven't adapted are struggling. The flexibility it offers—particularly for parents and caregivers—has been transformative. But it's not without its challenges—isolation, communication gaps, and the blurring of work-life boundaries—all remain real concerns."

Three sentences. Four em dashes. No human writes like this.

After (With ActMoreHuman's Em Dash Fix)

"Remote work has changed everything, from where we work to how we communicate. Companies that haven't adapted are struggling. The flexibility is real, especially for parents and caregivers. But so are the downsides: isolation, communication gaps, and the kind of blur between work and life that's hard to shake."

Same ideas. No em dashes. Reads like a person wrote it.

The ActMoreHuman Fix

ActMoreHuman solves this at the source. Instead of editing every AI response manually, you give your AI a persistent instruction to avoid em dashes entirely. It applies to every conversation going forward.

Our provider wizards generate a custom prompt that tells your AI exactly what to do—including killing the em-dash habit before it starts.

Fix It for Your AI Provider

Pick your provider below. The em-dash ban is already pre-configured:

Each link takes you directly to that provider's wizard with the em-dash rule already turned on. Run through the setup once, and it applies to every future conversation.

Other AI Punctuation Crimes

While you're in there, you might also want to address a few related patterns:

The "Certainly!" opener. AI models love starting responses with affirmations—"Certainly!", "Of course!", "Absolutely!"—that no human would ever say. ActMoreHuman eliminates these too. Bullet point everything. AI defaults to lists when prose would be better. The result is choppy, fragmented responses that feel like a PowerPoint deck instead of a conversation. The semicolon overload. Less common than em dashes, but semicolons show up at AI-scale frequency too; they get wedged in between clauses; in ways that feel stilted; because the model thinks it sounds intelligent.

All of these are addressable with the right system prompt. ActMoreHuman builds that prompt for you.

Frequently Asked Questions

Will this break other parts of the AI's responses?

No. Removing em dashes only affects punctuation. The AI's reasoning, accuracy, and helpfulness stay exactly the same. You just get the same quality of response without the mechanical giveaway.

What if I actually like em dashes sometimes?

Fair point. If you want em dashes in specific contexts, you can still type one manually. The instruction tells the AI not to reach for them by default—it doesn't prevent you from using them yourself.

Does this work for all providers?

Yes. The fix works for ChatGPT, Claude, Gemini, Perplexity, Grok, and Copilot. Each provider has slightly different setup steps, but the underlying principle is the same: a persistent system instruction that stays in effect for all future conversations.

Can I combine this with other humanizing instructions?

That's exactly what ActMoreHuman is designed for. The em-dash fix is one piece of a larger set of instructions that make AI writing feel genuinely human. Use it alongside our full humanizing prompt for the best results.

Ready to Stop the Em Dash Epidemic?

Pick your provider and fix it now. The wizard takes under a minute, and the em-dash ban applies to every conversation from that point forward.

Get started with the ActMoreHuman wizard and come back to this page if you want to configure other providers.

Why AI Loves Em Dashes—And How to Stop It

Ready to customize your AI?