AI Adoption and Audit Quality in Public Survey and Firm Inspection Data

Abstract

This study evaluates the association between AI-related questionnaire scores and audit-quality perceptions and examines what that association can establish about audit firms. A secondary analysis uses 98 respondent records and 25 items released by Handoko (2025). Equal-weight block scores are analysed using ordinary least squares, HC3 uncertainty estimates, rank correlations, alternative scoring, item-omission checks and row resampling. The AI–AQ correlation is 0.819. The AI coefficient is 0.735 in the unadjusted model and 0.731 after adjustment for four additional questionnaire blocks, with a 95% HC3 interval of 0.523–0.939. However, the AI–AQ heterotrait–monotrait ratio is 0.903, and several other block pairs exceed 1.00. Strong internal consistency therefore coexists with weak empirical separation. A separate descriptive analysis of UK Big Four inspection reports records 190 satisfactory or limited-improvement outcomes across 223 inspection instances in three cycles. These outcomes are not linked to the questionnaire respondents or to matched AI expenditure. The contribution is a reproducible assessment of the gap between observable association and an investment-effect claim. The evidence supports a positive association in the released questionnaire matrix, while neither its construct validity nor a causal return to firm-level AI expenditure is established.