OpenAI recently released data on how people use ChatGPT. One number stood out.
In December 2025, about 75% of messages from paid Pro users were work-related. For free users, the figure was about 27%. A year earlier, the gap between the two was roughly 25 percentage points. Now it is 48.
The pattern is clear: two large segments of usage are diverging around what is nominally the same product. One segment reaches for it when there is a work problem, the other reaches for it mostly outside of work, for conversation, reflection, or personal questions.
The obvious objection is that people who pay more naturally use it more for work. But ChatGPT Plus, at $20 a month, tells a different story: its work-related share has been falling alongside the free tier, not rising with Pro. Paying for the product does not explain the divergence. Something specific to the $200 tier does.
Whether that something reflects income, product design, or self-selection, this dataset cannot say. Probably all three, and the data cannot disentangle them.
Still, the gap has widened substantially over the past year, and it is not narrowing. If high-tier access correlates with work-intensive use, then access design becomes perhaps a distributional policy question, not just a pricing one. But that’s a question I leave to the economists.