How this report works

Sourcing standards

• Every figure is attributed to its original study, with the publisher, date, and — where available — sample size and population.
• We link to primary sources, not to other sites' coverage of them.
• We date everything. An adoption stat from 2024 is history, not news. Each page shows its last-updated date.
• We never blend figures from different studies into a single invented number.

Reconciliation standards

When credible studies disagree — and in this field they always do — we publish the disagreement rather than cherry-picking. For each conflicting set we identify: the definitionused ("any AI use" vs. "regular use" vs. "embedded in operations"), the population(all businesses, employer firms, customers of the surveying company), and themethod (probability sample vs. opt-in panel). See it in action: the adoption reconciliation.

A note on self-reported ROI

Statistics like "91% of AI users report revenue increases" are self-reported perceptions, not audited outcomes. Businesses that adopt AI also tend to be growing businesses, and respondents attribute gains generously. We publish these figures because they're widely cited — but we label them, and we weight government data more heavily than vendor data.

Update cadence

• Quarterly: full refresh of all statistics pages against the latest releases (Census BTOS, Fed, industry studies).
• As-published: major new studies are incorporated within days of release.
• Annually: our flagship report.

Corrections

Found an error? Tell us and we'll fix it fast, with a note. Independence and accuracy are the entire value of this project.