Every restaurant table in America holds the same two shakers. Most diners reach for them before tasting the food, a reflex so automatic that the chef’s intention disappears under a uniform layer of seasoning. Artificial intelligence has placed a third shaker on the table of human language, and it dispenses the em dash.
Anyone who reads professional content today has tasted it. The long horizontal line appears in LinkedIn posts, board memos, customer service replies, and quarterly letters to shareholders, sprinkled across sentences that never asked for it. The mark became so closely tied to machine writing that OpenAI CEO Sam Altman announced a fix in November 2025 that finally let users switch the em dash off, calling it a small win.
A punctuation habit sounds trivial to anyone running a company. What it reveals is far larger. The machine seasons every sentence the same way, and millions of professionals now consume that seasoning daily until it becomes their own taste.
The Vocabulary We Absorb
The em dash is the most visible ingredient, and the recipe runs much deeper. Researchers at the Max Planck Institute for Human Development analyzed more than 740,000 hours of academic talks and podcast conversations and found an abrupt rise in the words ChatGPT prefers after its release. Spoken use of “delve” climbed 48 percent and “adept” rose 51 percent within 18 months of the chatbot’s launch. The words surfaced in unscripted conversation as well, which means people were speaking them from memory instead of reading them from a screen.
The written record shows the same pattern. A study in Science Advances examined more than 15 million biomedical abstracts and concluded that at least 13.5 percent of 2024 abstracts were processed with large language models, a linguistic shift the authors found larger than the impact of the COVID pandemic.
This is subconscious programming at scale. Repetition is how the subconscious mind learns, and it accepts whatever it hears most often without asking where it came from. When a professional reads machine-shaped sentences dozens of times a day, those rhythms install themselves beneath conscious awareness and begin to feel like personal style.
The Constructions Hiding in Plain Sight
Vocabulary is only one layer. Large language models also favor a specific sentence architecture, and the most common is the contrast frame that tells the reader what something is not before revealing what it is. Writers now reach for this construction to manufacture insight where none exists, and readers have learned to feel its hollowness before they can name it.
The fluff arrives alongside it. Phrases about unlocking potential and fostering meaningful connections now coat corporate announcements like canned seasoning, applied regardless of what sits underneath. I keep a list of words I refuse to publish, and each one earned its place by losing all flavor through machine overuse.
Why the Palate Matters in the Boardroom
Leaders communicate through language long before they communicate through strategy. A CEO whose letters sound like every other CEO’s letters has surrendered the one asset no competitor can copy, which is a recognizable voice carrying a specific frequency of conviction. Stakeholders sense the difference immediately. When readers detect machine seasoning, they discount the message and question whether anyone at the top thought it through.
The flattening effect crosses borders. Cornell researchers found that when participants in India and the United States wrote with an AI assistant, their writing converged toward Western styles and lost cultural nuance, with Indian participants describing their own food and festivals through an American lens. For any organization operating globally, that finding is a warning about brand voice in every market it serves.
For years, the em dash was my shortcut. I dropped it into text messages in place of proper punctuation because it was easy and convenient, and I never gave it a second thought. When it began appearing across AI-generated writing, I recognized my own habit staring back at me, multiplied beyond reason.
The pattern that amuses me most is the contrast construction, the sentence that tells you what something is not before revealing what it is. This is the not X but Y style patterning. Every time I spot one, a red flag goes up that a machine shaped the words. What surprises me more is hearing it spoken aloud. People now use the same phrasing in meetings and conversations, and more than once I have caught it coming out of my own mouth.
Tasting Your Own Language Again
The remedy begins with awareness, because a pattern you can see loses its power over you. I recommend that leaders read one page of their own outgoing communication aloud each week and mark every rhythm they would never use in a live conversation.
Treat AI output as raw ingredients and do the seasoning yourself. Strip out the dashes and borrowed vocabulary, then rebuild each contrast frame into a direct statement of what you believe. Set explicit style rules for the tools your teams use, because the machine follows a clear instruction far better than it follows good intentions.
Language carries vibration, and the words a leader chooses set the frequency of an entire organization. The leaders who stand out in this decade will be the ones who taste their words before serving them to the people who trust them.
In my own kitchen, I have already made the switch. The salt and pepper now sit in the back of the cupboard, and chicken broth seasoning has taken their place on my table and in my writing, carrying a flavor that is recognizably mine. If you ever find an em dash in my work again, check the seasoning, because someone else has been cooking in my kitchen. And if a chatbot ever learns to season with chicken broth, I will know it has been reading my work.
