Beyond the Em Dash: Detecting and Avoiding AI Language in Corporate Communications

AI’s Preferred Sentence Structures Signal Assistance and May Weaken Impact 

11/20/26 – – We’re all getting better at detecting AI-generated language. Overuse of the maligned em dash is a recognized clue. And fingerprints turning up more and more in corporate prose are sentence structures grammarians call negative and antithetical parallelism:

It’s not X, it’s Y.

More than X, it’s Y.

It’s not only X, it’s Y.

This isn’t a problem of X, it’s a problem of Y.

Machines incorporate negative and antithetical parallelism in the language they generate because humans have used these rhetorical techniques effectively for centuries to make thoughts more memorable.

President Kennedy had no assistance from ChatGPT when he said, “Ask not what your country can do for you, ask what you can do for your country.” And it’s unlikely Shakespeare logged onto Claude to help him with, “I come to bury Caesar, not to praise him.”

But there can be too much of even a good thing. Expressing ideas with clarity and punch does not require contrast in every sentence. A simple declarative statement, especially in business communication, can do the job.

Consider this sentence: “Our financial performance was not only strong, we achieved the best quarterly earnings in our company’s history.” The first part of the sentence is unnecessary, delaying the momentous news of record results. “We achieved the best quarterly earnings in our company’s history” sounds good to me.

Remembering a Painful Editorial Lesson

A senior colleague of mine at the PR firm Hill and Knowlton tasked with reviewing junior account executives’ copy before it went to clients or journalists had his own way of discouraging unnecessary parallelism. He would write the bold letters “FF” (short for Forensic Foreplay) above introductory phrases he’d struck with his unmerciful red pen. “The reader doesn’t need a drum roll in every sentence before you make your point,” he would bark if you were unlucky enough to receive his editing “suggestions” in person.

That painful, valuable lesson comes back to me again and again as I view LinkedIn posts, listen to investor conference calls and read media releases.

Of course, you should not banish negative and antithetical parallelism from your rhetorical repertoire. Sometimes, especially in speeches and other oral presentations, a little forensic foreplay is appropriate. A TED Talk is different than an earnings release or conversation on Reddit.

5 Questions to Determine When Parallelism Makes Sense

Still, AI’s overuse of negative and antithetical parallelism (a recent Pew Research study found use has tripled in the last three years) is reason enough to ask yourself five questions before incorporating it in your communication:

  1. Is the sentence’s introductory, contrasting phrase relevant to making your primary point?
  2. Is the sentence more powerful without the qualifier ?
  3. Would the sentence be more efficient using a coordinating conjunction (and, or, but, yet, so)? “The quarterback is not only a passer, he’s a runner” can also be expressed as, “The quarterback can pass and run.”
  4. In how many sentences have you already used parallelism in what you’re writing?
  5. How many FFs would my former colleague have written on your draft?

To make sure what I’m observing is real, I asked ChatGPT. Here’s its answer:

“You are noticing a real linguistic phenomenon, and there is now fairly strong evidence that this particular contrast structure has become disproportionately common in AI-assisted writing.”

ChatGPT offered this defense:

“Language models aren’t simply trained to predict words. Modern systems are subsequently trained to produce answers that humans prefer.”

In other words: It’s not simply X, it’s also Y.

search previous next tag category expand menu location phone mail time cart zoom edit close