EUVIMEDEUVIMED entdecken
Crossref

Generative AI in Audit Oversight: A Checklist-Based Model Evaluation

Dejana Kresović, Sofija Drulović, Nebojša Đoković, Miloš Ćetković

IPSI Transactions on Internet Research · 01.07.2026 · 22 (3) · 38

DOI: 10.58245/ipsi.tir.2603.05 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IPSI Transactions on Internet Research, online erschienen am 01.07.2026. Verfasst von Dejana Kresović und 3 weiteren Autor:innen. Die Arbeit nennt 78 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper examines the use of generative artificial intelligence (GenAI) in audit oversight, with a focus on its role in the preliminary review of the independent auditor's report and accompanying financial statements. Existing audit supervision procedures rely on expert review, which is crucial but also time-consuming when it is necessary to examine large bodies of documents. The proposed solution is a framework supported by GenAI models, based on checklists, for identifying potential formal, structural and substantive problems in reporting that may require further expert review. The empirical analysis is based on 217 sets of documents of public companies in Serbia for the year 2024. ChatGPT and Gemini 3.1 Pro were evaluated the same documents using a structured checklist and a risk scale of 0 to 3, while the expert assessment of audit supervision served as a benchmark. Model performance was evaluated using summary risk scores, mean absolute error, bias, Cohen's kappa coefficient, precision, responsiveness, F1-measure, accuracy, and ROCAUC metrics. The results show that ChatGPT followed a more sensitive review pattern, with higher response and stronger adjusted agreement with the expert benchmark, while Gemini gave more conservative outputs, with higher precision but lower response. Both models performed better in formal and standardized checklist categories than in areas requiring professional judgment, such as key audit issues, business continuity considerations, and internal compliance red flags. The findings suggest that GenAI can support audit oversight by improving the consistency and prioritization of preliminary document review, but only when integrated into a structured framework with expert validation.

Zitieren

APA

Kresović, D., Drulović, S., Đoković, N., & Ćetković, M. (2026). Generative AI in Audit Oversight: A Checklist-Based Model Evaluation. IPSI Transactions on Internet Research, 22(3), 38. https://doi.org/10.58245/ipsi.tir.2603.05

Vancouver

Kresović D, Drulović S, Đoković N, Ćetković M. Generative AI in Audit Oversight: A Checklist-Based Model Evaluation. IPSI Transactions on Internet Research. 2026;22(3):38. doi:10.58245/ipsi.tir.2603.05

BibTeX
@article{Kresovi2026,
  author = {Kresović, Dejana and Drulović, Sofija and Đoković, Nebojša and Ćetković, Miloš},
  title = {Generative AI in Audit Oversight: A Checklist-Based Model Evaluation},
  journal = {IPSI Transactions on Internet Research},
  year = {2026},
  volume = {22},
  number = {3},
  pages = {38},
  doi = {10.58245/ipsi.tir.2603.05}
}
Kostenlos starten →