Financial auditing in the United States faces persistent challenges from escalating transaction volumes, complex financial instruments, and sophisticated fraud schemes that expose limitations in traditional sampling-based methodologies. This structured narrative review synthesizes peer-reviewed literature between 2020 and 2025, examining AI-enabled assurance frameworks that enhance audit accuracy, independence, and financial statement reliability. Methodological advances analyzed include machine learning for fraud detection, natural language processing for disclosure analysis, robotic process automation for continuous monitoring, and predictive analytics for risk assessment. Findings reveal AI systems substantially outperform conventional approaches in anomaly detection, timeliness, and coverage, with documented improvements of 40 to 50 percent in fraud detection speed and 30 to 40 percent reductions in audit cycle time. Persistent constraints include data quality limitations, algorithmic bias, cybersecurity vulnerabilities, interoperability challenges, and regulatory gaps under PCAOB oversight. Organizational barriers encompass auditor skepticism, skill deficiencies, and adoption resistance. Critical insights emphasize requirements for explainable AI, robust data governance, and preserved professional skepticism. Recommendations advocate comprehensive validation datasets, PCAOB-aligned AI standards, bias mitigation protocols, and clarified professional judgment roles to ensure AI augments rather than supplants human expertise in U.S. financial auditing.