US daily AI use more than doubled, but the survey shifted

What the surveys measured

The March survey included 2,017 people; the August survey included 1,016. Both samples were randomly selected and weighted to represent the US population.

In March, respondents were asked how often they used AI in general. In August, they reported usage separately for each service, and researchers used the highest frequency they gave.

The number needs a qualifier

Epoch AI says the results cannot be fully compared. The August method may actually understate how often people use AI, even though the headline result shows daily use more than doubling over the six-month period.

I think the methodological mismatch is the central fact here. A rising number could reflect more frequent use, broader exposure to AI services, or simply the decision to measure each service separately and then keep the highest reported frequency. The survey supports a clear signal of rapid change, but not a perfectly measured growth rate.

The question the announcement leaves open is how much of that increase survives when the same question is asked in the same way. Until that comparison exists, the doubling is best read as evidence of momentum, not a precise map of US AI habits.