Generative Artificial Intelligence in Accounting and Auditing: A Bibliometric and Thematic Review

This paper aims to map the knowledge structure of the research on the generative artificial intelligence and the large language models in accounting and auditing, and to synthesise the previous findings into an integrative framework and a research agenda. A bibliometric analysis was combined with a thematic content analysis. The records were retrieved from OpenAlex and Crossref for the period from January 2022 to April 2026 and screened following the PRISMA 2020 logic, and the final corpus consists from 321 peer-reviewed documents. The science mapping was conducted through a concept co-occurrence network and the Louvain clustering. The research grew in an exponential manner, where 75.4% of the corpus was published in 2025 and 2026. The United States and China dominate the production, while only 16.5% of the documents were internationally co-authored. The citations are concentrated, since the ten most cited documents received 47.7% of the citations. Four clusters were identified, namely the reporting, disclosure and markets, the education, acceptance and the profession, the judgment and model reliability, and the auditing and assurance. The evidence moved from the benchmarking of the chatbots on the examinations toward the evidence on the audit quality, the disclosure and the market efficiency, but the causal and longitudinal evidence is still limited. The study proposes an enablers, mechanisms and outcomes framework and a research agenda for the researchers, the regulators, the audit firms and the educators, especially in the emerging economies.