AI writing tells

Uniform sentence rhythm

Read a page of AI text out loud. Somewhere around the fourth sentence you'll feel it, even before you can name it. Every sentence takes the same number of breaths. Fifteen to twenty words. Subject, verb, object. Again. A metronome that happens to produce grammar.

Researchers call the opposite quality burstiness, which is a silly word for something you already know: people write in jabs and sprawls. A three-word sentence here. Then a thirty-five-word thought that wanders off, doubles back, and lands somewhere you didn't expect it to, possibly mid-shrug. Language models can't help regressing toward the statistical middle of all that. And the middle, it turns out, is a seventeen-word sentence.

Why detectors flag it

Burstiness is one of the two core signals AI detectors measure (the other is perplexity, or how predictable your word choices are). It's cheap to compute (count words per sentence, look at the variance) and hard to fake accidentally. A low variance across a whole post is strong evidence no human rhythm was involved.

What it looks like

Remote work has changed how teams communicate every day. Managers need to be intentional about creating connection points. Regular check-ins help people feel included in decisions. Small rituals can make a distributed team feel cohesive.

Four sentences: 9, 9, 9, and 10 words. Nothing is wrong with any one of them. Together, they're a metronome.

How to fix it

• Count, don't guess. Word-count your sentences. If most land within a few words of each other, break the pattern deliberately.
• Add a short one. Three to six words. ("That's the whole trick.")
• Let one run long. A sentence that stacks a real thought on a real thought and doesn't apologize for taking its time to get where it's going.
• Vary the openings. If every sentence starts with the subject, rhythm stays flat even when length varies.