The Surprising Insights: How AI Assistants Recommend CPA Firms for Auditing

Unveiling the Flaws in AI Recommendations for CPA Firms

In a groundbreaking study conducted by AIWorthy, the manner in which AI assistants recommend CPA firms for federally mandated audits has come under scrutiny. The research revealed startling outcomes regarding the efficacy and reliability of AI in this critical decision-making process. As organizations strive to maximize their expenditures on professional services, understanding how artificial intelligence makes these referrals is more important than ever.

The study encompassed an evaluation of prominent AI assistants, including ChatGPT, Claude, Gemini, Perplexity, and Google AI Mode. These platforms were posed ten questions each, mimicking the inquiries of a first-time buyer with the intent of selecting a suitable CPA firm for a Single Audit, a requirement for those organizations expending $1 million or more in federal funds annually. Throughout this exercise, the AI assistants provided answers on two separate dates - August 11 and September 28, 2026.

Findings That Challenge Conventional Wisdom

One of the most significant revelations was that the firms actively engaged in Single Audits were seldom recommended. The four most busy CPA firms, collectively managing 45 out of the recommended 75 federal engagements, were only referenced 24 times among 250 responses generated in August. In stark contrast, the six smallest firms, responsible for a mere 30 engagements, were suggested 184 times. Alarmingly, three firms that led the field in workload went unmentioned entirely during the recommendation process on both assessment dates.

The research highlighted three distinguishing characteristics shared by the firms that received the most recommendations. They all possessed an updated homepage specifically showcasing their involvement in the Triangle area and a detailed description of the type of organization they cater to. Conversely, firms that were never recommended lacked at least one of these crucial elements. Disturbingly, the busiest firms failed to meet any of them, raising concerns about how the relevance of their service was recognized by AI.

The Glaring Absence of Source Verification

A notable aspect of the study was the rare citation of public records by the AI assistants. Among the two that managed to reference sources, they collectively acknowledged 2,725 pages; however, the Federal Audit Clearinghouse, which catalogues every auditor engaged in Single Audits, was cited merely four times. Curiously, not a single mention was made of the state licensing board. This neglect towards public records casts doubt on the authenticity of recommendations provided by AI.

Moreover, the randomness of recommended firms extended to instances where buyers were directed towards names of firms that no longer exist. There were 79 instances in August and 85 in September where AI suggested agencies under outdated names, highlighting a significant gap in the AI systems' ability to keep up with current data. One notable example included a firm whose previous name was retired in April 2024. Such recommendations risk undermining buyer confidence and could lead to lost opportunities for effective collaboration.

Adding to the peculiarities was the case of a frequently recommended firm, noted 78 times in the AI answers, that had a website still under construction. For prospective clients, a maintenance page offers little in terms of trust or information about capabilities.

The Call for Transparency in AI Recommendations

According to Heather Harreld, the founder of AIWorthy, “Firms are being asked to invest in AI visibility without an independent assessment of the methodologies used by assistants when answering buyers’ requests.” This raises critical questions regarding the accountability of AI platforms in providing accurate recommendations. Understanding what AI gathers, how it interprets information, and ultimately, the choices it suggests should be pivotal for any firms considering budgeting for AI services.

The panel used in the study ensured that questions and scoring were firmly established through cryptographic hashes before any firms were contacted, safeguarding the integrity of the results. Every CPA firm assessed received transparency regarding their recommendations alongside an opportunity to respond. However, no firm names were disclosed in the findings, maintaining their anonymity amidst the investigative process.

This thorough analysis paves the way for further discussions about the necessity of AI accountability frameworks in the B2B sector. As businesses increasingly turn to artificial intelligence for recommendations, ensuring the accuracy, relevance, and integrity of such information should remain a prerequisite in potential partnerships.

For further reading, the complete study and visual summary can be accessed at AIWorthy’s official site. This research could serve as a stepping stone for organizations to reassess their engagement with AI platforms and focus on obtaining transparent, reliable recommendations in order to make informed business decisions.