Decomposing Germany's Low Reported Student Use of Generative AI for Formal Education: Evidence from the 2025 EU ICT Household Survey

In the 2025 EU ICT household survey, 54.6% of EU27 students reported using generative AI for formal education in the past three months, against 29.1% in Germany, the lowest of 35 European countries. It asks whether German students use generative AI less overall or report the education purpose less often, and whether Austria shares the pattern. The analysis follows a plan with one outlier rule (below the first quartile minus 1.5 IQR of the other countries) and five stop conditions; its timing relative to the first-pass results cannot be verified, so it is not called pre-registered. The planned premise failed: Germany's 29.1% lies above the low fence of 7.3%, so the paper does not claim Germany alone is low. For reference, the planned decision table assigns (B) a low education share, then (D) Germany only: overall use among German students (64.8%, reference median 79.7%) is low but not an outlier, while the share of student users naming formal education, approximated by FE/U, is 0.45 against a median of 0.79 and a low fence of 0.53 (outlier low, lowest of 34); the private-purpose share (0.78) and Austria (0.83) are normal. The rule does not apportion the gap between them. Outside the plan's wording rule, the German questionnaire's second purpose option is ``job-related education and training'' where the Eurostat model has ``professional (work) purposes''. In exploratory checks German students report it at 26.9% (reference median 11.0%), German employed people at 20.8% (median 26.1%, normal range). Purposes are multiple response with unpublished overlap, so the share of student users naming either option lies within 0.45 (lowest of 34) to 0.86 (normal); whether the two options together close the gap is unknown. This fits some German students filing study-related use under the second option, but the data cannot test that cause, and Eurostat publishes no standard errors.