The Em Dash Has Somehow Become an AI Disclaimer

dehakuran.com · August 2026 · 2 min read

The em dash has somehow become an AI disclaimer.

People are suddenly removing certain punctuation, changing their tone, adding little imperfections — all to prove: “I wrote this myself.”

Funny how fast we got here.

The Authenticity Flip

Not long ago, people were proudly sharing AI-generated portraits of themselves. Now the conversation has flipped to detection, watermarking, and proving that content is “authentic.”

Anthropic saying it will watermark AI-generated text tells us something about where this is heading.

Two Assumptions

Maybe two assumptions are forming at the same time:

  1. AI companies think AI is becoming good enough to pass as human.
  2. Humans think AI is becoming good enough to replace them.

I’m not convinced by either.

AI Did Not Create the Problem

Authenticity was already a problem long before AI. AI didn’t create performative expertise, fake confidence, recycled opinions, or content written mainly for algorithms.

It just industrialized them.

And perhaps that’s the bigger risk: not that AI becomes indistinguishable from human intelligence, but that human communication gradually optimizes itself to the level of machine-generated content.

The problem may not be artificial intelligence becoming too intelligent. It may be societal intelligence becoming more artificial.

And yes — I used an em dash and AI to generate an image 🤘

Frequently Asked Questions

Why has the em dash become associated with AI-generated writing?

Because readers began treating recurring stylistic signals—especially polished punctuation and consistent tone—as evidence of AI assistance. This has created a strange authenticity ritual: people now remove punctuation and introduce imperfections to prove, "I wrote this myself."

What does Anthropic's text watermarking signal?

It signals that AI provenance is becoming a product and trust problem, not merely a writing-style debate. Anthropic's move toward machine-detectable watermarking reflects growing demand to identify whether content was generated or processed by AI.

What is the deeper risk of AI-generated content?

The deeper risk is not simply AI becoming indistinguishable from humans. It is human communication optimizing itself toward the polished, repetitive, algorithm-friendly patterns associated with machine-generated content—making societal intelligence more artificial.

AI StrategyFuture of WorkTech Trends

Deha Kuran

AI Executive, Engineer, and Evangelist. Head of AI Business Operations at Philips.

Follow the thinking on LinkedIn →