Writers Are Deliberately Writing Worse to Prove They Are Human

Writers who produce all their own work are now going out of their way to sound less polished, adding casual language, intentional typos, and run-on sentences to signal human authorship to a growing audience of amateur AI detectors. The Wall Street Journal reported on the phenomenon, describing it as a reverse Turing Test playing out across Substack, LinkedIn, and social media.

The tells that have emerged are specific. Heavy use of em dashes, staccato sentence cadence, lists of three, and phrases like "it's not x, but y" are now widely associated with AI output. One financial account coordinator interviewed in the piece said he removes em dashes from his own AI-assisted writing and deliberately leaves in run-on sentences to retain the sound of his own voice. A tech founder replaces em dashes with two smaller dashes on his LinkedIn posts. Others are scattering obscure references and colloquialisms that a model would be unlikely to generate.

The irony flagged by one startup founder in the AI-humanizing space: text rewritten by humans is increasingly being flagged by AI detectors, suggesting people are unconsciously starting to copy AI's style.

Head of Financial Narrative, Ashley Jones, notes in a recent contribution to Tearsheet, that financial communications professionals have already been grappling with this question: where AI belongs in the content process and what gets lost when the thinking is skipped entirely.

Full story: WSJ