AI writing tells

Parallel paragraph structure

Look at an AI-written post's paragraphs side by side and you'll often find the same skeleton in each: a setup sentence, an observation, a tidy evaluative close. Three sentences, self-contained, complete with its own miniature arc. Then the next paragraph does it again. And again.

Human paragraphs are unruly. Ideas spill across paragraph breaks. Some paragraphs are one sentence. One is just a question. A thought started in paragraph two gets picked back up in paragraph five, because that's how remembering works.

Why detectors flag it

Structural self-similarity is measurable: paragraph lengths, internal sentence roles, and the absence of cross-paragraph callbacks. Models default to parallel structure because each paragraph is locally optimized to be complete. Completeness, repeated four times, is the tell.

What it looks like

Async standups saved us time. We reclaimed five hours a week per engineer. That time went straight back into deep work.

Written updates improved clarity. People think harder when they write. The quality of our decisions went up.

Fewer meetings helped morale. Nobody misses status theater. The team is happier for it.

How to fix it

• Merge two paragraphs and let one idea interrupt another, the way it did in your head.
• Make one paragraph a single line.
• Call back. Reference something from earlier ("that Friday email again") instead of marching strictly forward.
• Leave one thought unfinished and pick it up later. Or don't.