Can ChatGPT be skeptical? A case-based evaluation of generative AI in financial statement auditing

Purpose This study aims to investigate how effectively GenAI (ChatGPT) displays professional skepticism characteristics in external audit tasks, as insufficient skepticism may impair audit quality and heighten the risk of audit failure. Design/methodology/approach ChatGPT is evaluated using four audit tasks representing common judgment-oriented audit procedures. Three experienced external auditors independently assessed ChatGPT’s responses using a structured professional skepticism assessment. Findings ChatGPT appears to display varying levels of skepticism characteristics across audit tasks, with relatively higher levels in preliminary risk assessments and substantive analytical procedures and lower levels in going-concern evaluations. This suggests that it may be more effective in analytical tasks than in those requiring deeper professional judgment. Originality/value This study appears to be among the first to systematically assess whether a general-purpose reasoning GenAI tool can display characteristics associated with professional skepticism across different audit tasks. By applying an established skepticism framework, it provides a structured baseline for understanding where GenAI may support or fall short in audit judgment contexts.