Habr AI explained why the em dash became a marker of AI-generated text

On Habr AI, an article was published about an unexpected side effect of the mass use of neural networks: an em-dash, careful typography, and overly polished editing have begun to be perceived as signs of machine-generated text. The author argues against this habit and shows why grammatical correctness should not become something suspicious.

Why the em-dash frightens

Not long ago, an em-dash was for an editor almost an automatic marker of quality: a person understands the difference between a hyphen, an en-dash, and an em-dash, monitors spaces, and does not leave typography to chance. But after the boom of generative models, this same precision began to work against the author. Readers and clients increasingly search not for semantic errors, but for stylistic patterns associated with ChatGPT and other systems, and the em-dash unexpectedly ended up on the list of such suspicious signs.

"Today this is a red flag of neural networks."

Because of this, a new oddity emerged: people deliberately simplify the text, replace the em-dash with a hyphen, remove typographic habits, and even allow roughness, just so the material looks more human. In the article, this shift is shown not as a minor editorial fashion, but as a symptom of a broader problem. If the audience begins to confuse grammatical correctness with machine generation, it is not AI that suffers, but the very norm of written speech, which was previously considered an advantage of the author.

When to use an em-dash

The material on Habr AI is not limited to cultural observation and quickly moves to practice. The author reminds us that the em-dash in Russian is needed not for beauty, but because it has specific functions: it separates parts of an utterance, sets a pause, marks an omission of a connecting word, and helps the reader instantly grasp the structure of a phrase. Therefore, the habit of replacing it with a hyphen for fear of being accused of using AI leads not to naturalness, but to a loss of clarity.

• distinguish a hyphen within a word from an em-dash between parts of a sentence
• place an em-dash where it marks a pause or logical opposition
• remember the spaces around the em-dash in standard Russian typography
• use keyboard shortcuts to avoid spending time on manual correction

A separate useful part of the text is a small guide to quickly typing an em-dash. This makes the material not just a column about linguistic anxiety, but a practical tool for editors, authors, social media specialists, and everyone who writes in Russian every day. The logic is simple: if a sign is needed by the rule, it should be used without regard to fashionable fears. Otherwise, the text begins to mimic carelessness, although the task of editing is the exact opposite.

The editor and AI

Toward the end, the author moves from punctuation to a broader question: how to use neural networks in editorial work so that they empower humans rather than replace them. The answer is practical and therefore convincing. AI is convenient where you need to quickly put together a draft, suggest wording options, reduce routine work, or test several approaches to presentation.

But the final decision still rests with the editor, because only a person understands the context, intonation, target audience, and the boundary where helpful assistance turns into mechanical writing. This conclusion is particularly important against the background of a growing fashion to hide any traces of good editing. If an author deliberately worsens the typography to avoid being suspected of working with a model, they lose twice: they lose precision and play into a weak criterion for evaluating text.

A much more productive approach is different: use AI as a draft assistant, while leaving quality, correctness, and typographic cleanliness in the responsibility of humans. Then technology saves time, but does not blur the professional standard.

What this means

The story with the em-dash well illustrates how quickly cultural markers around AI change. The problem is not the neural network itself, but the habit of evaluating text by superficial signs instead of meaning, structure, and precision. For media, editors, and brands, the conclusion is direct: do not imitate errors for the sake of apparent humanity. It is much more important to establish a proper process in which AI accelerates the work, and humans are responsible for the quality of language and trust in the text.