Demand
U.S. Business AI Use
Weighted share of U.S. nonfarm employer businesses reporting AI use in the prior two weeks—a timely gauge of how broadly AI is diffusing into business operations. Census broadened the question in November 2025, creating a separate series that must not be spliced to the original.
U.S. Business AI Use
In the latest BTOS period ending 2026-08-23, the revised current-use question measured 22.4% of U.S. employer businesses. Do not compare this level directly with the original-wording series.
Methodology
Read the national Yes row for Question ID 7 from the official AI Core Questions historical workbook and the current National workbook, pairing each percentage with its matching official Response Standard Error and collection-date column.
Retain every published biweekly percentage exactly as shown. Exclude dots used by Census for unavailable periods; do not interpolate gaps or calculate forecasts.
Original exact question: “In the last two weeks, did this business use Artificial Intelligence (AI) in producing goods or services? (Examples of AI: machine learning, natural language processing, virtual agents, voice recognition, etc.)”
Revised exact question: “In the last two weeks, did this business use Artificial Intelligence (AI) in any of its business functions? (Examples of AI: machine learning, natural language processing, virtual agents, voice recognition, etc.)” Census says the revised wording captures uses such as hiring, project management, accounting, R&D, and AI embedded in software workflows that respondents could omit under the original wording.
Treat sample period 202524 (collection November 17–30, 2025; released December 4) as the first observation of a separate series. Do not calculate growth across the break. Sample periods 202521–202523 are documented structural gaps.
BTOS is a voluntary, nationally representative survey of U.S. nonfarm employer businesses. Published point estimates remain subject to sampling error, coverage and measurement error, processing error, and possible nonresponse bias. The official standard error is retained in each CSV note; this line chart does not show an uncertainty band.
Frequently asked questions
What exactly does this chart measure?
The weighted percentage of in-scope U.S. employer businesses that answered Yes to the BTOS question about using AI in the last two weeks. It is not a share of employees, spending, output, or usage frequency.
Why does this metric matter for the AI industry and economy?
It is a high-frequency measure of business diffusion: whether AI has moved from suppliers and pilots into active use by employer firms. It helps distinguish broad operational adoption from model releases, investment, spending, or employee-level exposure, none of which this survey directly measures.
How should the two wording eras be interpreted?
Read movement within each line, not across the break. The original question covered AI used to produce goods or services; from sample period 202524, the revised question covers any business function. Census observed a level shift and designated the December 4, 2025 release as a new time series, so the apparent jump is not sudden adoption growth.
Where do the values come from, and are they modeled?
Values are official weighted Yes estimates for Question ID 7, Answer ID 1 from the Census BTOS national workbooks, matched to the published collection period and standard error. No value is forecast, interpolated, smoothed, or otherwise modeled.
What are the main survey limitations?
BTOS is a voluntary probability survey of U.S. nonfarm employer businesses. Estimates are subject to sampling uncertainty, nonresponse bias, coverage, measurement, and processing error. Point-specific standard errors are retained in CSV notes, but the chart does not display an uncertainty band.
Why is there a gap in autumn 2025?
Census did not collect BTOS responses for the three periods covering October 6 through November 16, 2025 and classifies them as structural gaps.