LLMs love the em dash — ask AI to write something and you'll find several scattered throughout.

Most people have learned to spot this pattern, so the em dash now triggers unconscious devaluation. Even when we rationally know AI can write well, and that the author's judgment matters more than the tool, something makes us automatically think "this is worse" once we suspect it's written by AI.

But sometimes the em dash really is the best choice. Use it, though, and you risk that instant reader dismissal — even if your writing is genuinely good.

You're left choosing between two camps: write what works best and ignore reader bias, or prioritize reader perception and avoid the unconscious discount.

I lean toward the first camp. Good writing shouldn't have to hide from its own tools just because algorithms overuse them. But if you're trying to persuade someone who might not know you well? Maybe save the em dashes for when you've already earned their trust.

LLMs love the em dash — ask AI to write something and you'll find several scattered throughout.

Most people have learned to spot this pattern, so the em dash now triggers unconscious devaluation. Even when we rationally know AI can write well, and that the author's judgment matters more than the tool, something makes us automatically think "this is worse" once we suspect it's written by AI.

But sometimes the em dash really is the best choice. Use it, though, and you risk that instant reader dismissal — even if your writing is genuinely good.

You're left choosing between two camps: write what works best and ignore reader bias, or prioritize reader perception and avoid the unconscious discount.

I lean toward the first camp. Good writing shouldn't have to hide from its own tools just because algorithms overuse them. But if you're trying to persuade someone who might not know you well? Maybe save the em dashes for when you've already earned their trust.