Professional skepticism is a core requirement for auditors using AI tools. Because AI can suffer from " hallucinations " or misinterpretation of data, the auditor must " Validate the evidence against source documents. " This involves taking a sample of the AI ' s findings and manually verifying them against the original records (e.g., invoices, contracts, or logs). According to the ISACA AAIA™ Study Guide, while reviewing model cards (Option A) provides context on the tool ' s capabilities, it does not confirm that a specific piece of evidence is accurate. Direct verification against the " Ground Truth " (source documents) is the only way to ensure the reliability and integrity of AI-generated audit evidence.