Atlantic tracks AI's 'not X, but Y' tic into Fortune 500 filings

Every model's writing has a tell now, and The Atlantic argues the most famous of them is the "it's not X, it's Y" construction, the antithesis chatbots reach for whenever they want to sound thoughtful. The magazine frames it as also the most mysterious of AI's stylistic habits, well known enough that once you start noticing it, you see it everywhere.

The mechanics are older than the models. In classical rhetoric this is antithesis, or negative-positive parallelism, and it works by first negating a familiar assumption then replacing it with something supposedly more expansive. Shakespeare gave the archetype in Julius Caesar: "The fault, dear Brutus, is not in our stars, but in ourselves." Human writers use the move as a high-impact device deployed sparingly, because the whole effect depends on restraint. Language models do not have that governor, so they apply the structure whether the subject is a recipe, a bug fix, or a business report.

The scale is now measurable. The Atlantic cites a Washington Post dataset in which variations of "not just X, but Y" appeared in roughly 6% of all July messages, an unusually large share for a single rhetorical shape. It also cites Barron's tallying the same construction in Fortune 500 filings, where it climbed from 50 in 2023 to over 200 in 2025. The reporting adds that cognitive psychologists have tested this shape since at least 2003 and found that readers process the negated noun first, so the framing may land less crisply than the writer intends even when a human is the one using it. Four experts in our Who's Who directory shared the piece, which tracks with how quickly the pattern has become a running joke among people who read a lot of model output.

The gap the piece leaves open sits right in its headline. The Atlantic calls the tic mysterious because it does not offer a mechanistic answer for what in pretraining or fine-tuning pushes models so hard toward this shape, and it does not resolve whether prompting a model away from the construction actually suppresses it or just relocates it. Read the frequency counts as the reporting has them, not as a closed case.

For anyone drafting a shareholder letter, an earnings script, or a launch blog, the takeaway is the smaller one. The construction has flipped its polarity in the reader's mind. Overused, it advertises machine authorship rather than emphasis, and stripping it from a draft may be the cheapest voice edit a comms team can make this year.