# Methods and limitations

## EU extraction

On 20 October 2026, Ariftly queried Eurostat dataset `isoc_eb_ai`, updated 15 June 2026. The selected dimensions were `geo=EU27_2020`, `unit=PC_ENT`, `indic_is=E_AI_TANY`, the standard activity aggregate `C10-S951_X_K`, and the relevant year and size classes.

The API returned 19.95% overall, 17.00% for 10–49 people employed, 30.36% for 50–249 and 55.03% for 250+. Eurostat's publication rounds these to 20.0%, 17.0%, 30.4% and 55.0%. The 2023 and 2024 overall cells were 8.06% and 13.48%, published as 8.1% and 13.5%.

Eurostat's population excludes enterprises below ten people employed and excludes some economic activities. Its 2025 indicator includes a new category covering generation of pictures, video, sound or audio. Trend calculations therefore carry a measurement-change note.

## U.S. extraction

The 18% firm-weighted and 32% employment-weighted estimates are taken from the abstract of U.S. Census working paper CES-26-25. Its stated supplement reference period is November 2025–January 2026. The figures are alternative weightings of the same underlying adoption concept, not separate surveys.

The core BTOS continues to publish biweekly waves. The article uses dated wave examples only when their period is stated; it does not relabel the working-paper supplement estimate as the latest BTOS rate.

## Comparison contract

Direct arithmetic is permitted only within a marked comparison group. The EU size values may be compared within the 2025 Eurostat population. The U.S. firm- and employment-weighted values may be contrasted within CES-26-25. The EU and U.S. headline values are `descriptive_context_only`: their denominator, industry coverage, size threshold, question and period differ.

No confidence intervals were derived. Ratios and percentage-point changes use published rounded values and must not imply additional precision or causation.