Artificial intelligence is no longer an occasional tool for most Americans — at least according to the latest survey evidence. Research institute Epoch AI, working with polling firm Ipsos, has released findings from two representative surveys showing that near-daily AI use in the United States rose sharply over a six-month stretch in 2026.
The change is large enough to read as a genuine shift in behaviour rather than statistical noise. Between March and August 2026, the proportion of U.S. adults who reported using AI on at least six out of seven days climbed from 8 percent to 19 percent — better than a doubling in less than half a year.
What the Two Surveys Measured
Both waves of the study were fielded by Epoch AI with Ipsos, and both were built around samples that were randomly selected and weighted to mirror the U.S. population. The first wave, conducted in March 2026, drew on 2,017 participants. The second, fielded in August, included 1,016 participants.
The headline comparison rests on a single question of frequency: how many days out of a typical week a person turns to AI. The answers show a population moving from sporadic experimentation toward something closer to routine reliance.
The Numbers at a Glance
- The share of U.S. adults using AI on at least six of seven days rose from 8 percent in March 2026 to 19 percent in August 2026.
- The share of occasional users — people who engaged with AI just one day a week — fell from 17 percent to 10 percent over the same period.
- The March survey included 2,017 respondents; the August survey included 1,016, with both weighted to represent the U.S. population.
Taken together, those two movements describe a population that is not simply adding AI to the list of things it occasionally tries. The heavy-use group expanded while the light-use group contracted, which points to deepening engagement among people who were already encountering these tools.
A Retreat From Occasional Use
The decline in weekly dabblers may be the more revealing half of the story. A rise in near-daily use could, in theory, reflect a wave of first-time users who immediately became habitual. But the simultaneous drop among people who touched AI only once a week suggests something else is happening: existing casual users are moving up the frequency ladder.
That distinction matters for anyone trying to forecast where adoption goes next. If the growth came purely from new entrants, the trend might stall once the pool of curious newcomers thins. If instead the movement is driven by people already familiar with the technology using it more often, the ceiling on near-daily usage is considerably higher than the current 19 percent.
The surveys do not resolve which explanation dominates, but the falling share of one-day-a-week users is consistent with the second pattern — a gradual deepening rather than a one-off surge of novelty.
How the Data Was Collected
Epoch AI is a research institute that has built a reputation for quantitative work on artificial intelligence, and the collaboration with Ipsos gives the findings the machinery of professional opinion polling: random selection, population weighting, and sample sizes in the thousands for the first wave.
Weighting matters here. Without it, a survey of roughly 2,000 people could be skewed by the demographics most likely to respond — typically younger, more online, and more technologically inclined respondents. Adjusting the sample to match the country as a whole is what allows the resulting percentages to be discussed as estimates of national behaviour rather than descriptions of a self-selected group.
A Caveat That May Understate the Growth
The researchers are upfront that the two waves are not perfectly comparable, and the reason is a change in how the question was asked. In March, participants were asked how often they used AI overall. In August, they were asked about each service separately, and the researchers then took the highest frequency reported across those services.
That methodological difference could cut in an important direction. Asking service by service may lead respondents to overlook some of the tools they use, or to under-report the total number of days on which AI played a role in their work or personal life. If that is what happened, the August figures would understate real usage rather than exaggerate it — meaning the measured jump from 8 percent to 19 percent could be a conservative reading of the change.
Either way, the discrepancy is a reminder that survey trends of this kind are estimates with known seams, not exact census counts. The direction of travel, however, is consistent across both waves.
Why the Numbers Matter
Frequency of use is one of the clearest signals of whether a technology has moved from curiosity to infrastructure. Products that people open once a week occupy a very different place in their lives — and in the economy built around them — than tools they consult almost daily.
A near-daily habit has knock-on effects that reach beyond the software itself. It shapes how people search for information, draft documents, solve problems, and make decisions. It also changes expectations: once a workflow depends on AI assistance several days a week, giving it up imposes a real cost, and that dependency tends to be sticky.
For companies building AI products, the finding that the casual-user tier is shrinking is arguably as important as the growth at the top. Retention and habit formation, not just acquisition, appear to be driving the numbers.
Open Questions the Surveys Leave Behind
Frequency tells us how often people show up, but not what they do when they get there. These surveys do not break down near-daily usage by task, by industry, or by the specific services involved — the August methodology asked about services individually, but the published comparison centres on the highest reported frequency per person.
Nor do the results explain the underlying cause of the shift. The surveys capture behaviour, not motivation, so the jump could reflect improving tools, growing familiarity, workplace expectations, or some combination of all three. Distinguishing among those explanations would require a different kind of study.
What to Watch Next
The obvious next datapoint is a third wave. Another survey using the August question format would settle part of the comparability problem, showing whether near-daily usage continues climbing or plateaus now that it has reached roughly one in five adults.
It is also worth watching whether the occasional-user share keeps falling. If one-day-a-week usage continues to shrink while the near-daily group expands, the picture becomes one of consolidation — a technology settling into the rhythm of ordinary life rather than spiking and fading. The six months from March to August 2026 already look like a meaningful step in that direction.
This article is based on reporting by The Decoder. Read the original article.
Originally published on the-decoder.com
