CPAs vs. AI Tax Advisors: Evaluating Reliance Across Advisory Contexts

Abstract

Advancements in generative artificial intelligence (AI) have led to the public deployment of AI chatbots that provide tax planning and compliance advice. This monumental development in the tax advisory landscape enables individual taxpayers who prepare their own tax returns to meaningfully conduct their own tax research. In this study, I address two fundamental questions. First, to what extent are taxpayers willing to rely on AI advisors? Reliance on AI is compared to that placed on certified public accountants (CPAs), a known benchmark for tax advice quality. Second, this development increases the importance of understanding taxpayers’ preference for aggressive or conservative advice, particularly for those who prepare their own tax returns. Do self-preparers exhibit a preference for aggressive advice? To investigate these questions, I conduct a randomized experiment that manipulates advisor type and aggression to shed light on how these factors influence taxpayers during the reporting process. Trust and advocacy are examined as mechanisms driving differential reliance on advisors. Results provide evidence that self-preparers are influenced by AI advisors, and are more likely to rely on AI-provided advice when it is aggressive than when it is conservative. However, taxpayers express less reliance on AI compared to CPAs, an effect driven by perceptions of CPAs’ tax expertise and commitment to minimizing tax burden. The study provides insights for the research community, professional tax service providers, legislators, regulators, and other policy makers considering the implications of this rapidly evolving technology.