IRS Uses AI to Help Select Tax Returns for Audits

The Internal Revenue Service (IRS) is now using artificial intelligence (AI) to help decide which tax returns might need a closer look. This technology helps the agency sort through millions of filings to find those that seem most likely to have issues. While AI plays a role in ranking these returns, human reviewers still make the final decisions about who gets audited. This ensures that important choices are made with human judgment.

The IRS considers AI systems that influence audit selection as “high-impact.” This means they are part of the agency’s most sensitive technology use. The goal is to use data and advanced technology to focus compliance efforts where they are most needed. This approach builds on older methods the IRS has used for years to analyze tax data.

How AI Helps Prioritize Tax Returns

The IRS has been expanding its use of AI and data analytics. As of March 2024, the agency had 126 AI use cases, with many developed recently. In tax compliance, AI is used to help select returns for audits and to find errors or fraud. This technology helps speed up the process and identify patterns that might be missed otherwise.

The IRS has tested AI models for case selection, spending about $1.8 million on one contract. These tools are designed to rank risk and identify complex or high-risk returns. While AI can quickly sort and prioritize filings, it does not make the final audit decision on its own.

Traditional Audit Triggers Remain Important

Even with AI, many of the reasons a tax return might attract IRS attention are still the same. These include having a high income, filing complex returns, claiming aggressive tax credits, or taking unusual deductions for charities or business expenses. AI helps to sort through these factors more quickly and efficiently across a large number of tax filings.

For example, a complex return with unusual deductions would likely already be flagged for review. AI can help determine how high up that return moves on the list for examination and which specific parts of the filing should be examined more closely. It acts as a tool to enhance the existing screening process.

Addressing Concerns About Bias

There are ongoing concerns about whether AI systems might unintentionally create bias in audit selection. Past studies have shown that certain groups, like Black taxpayers, have been audited at higher rates. The Government Accountability Office (GAO) has also pointed to “unintentional algorithmic biases” as a potential issue.

While this does not mean the IRS deliberately discriminates, it highlights the need for careful oversight of automated systems. Because AI can influence who gets audited, lawmakers and watchdogs are pushing for clearer rules on how these systems are tested, documented, and reviewed. The IRS’s classification of these systems as “high-impact” shows they recognize the need for oversight.

Human Review Remains Key

It is important to understand that the IRS has not fully automated audit decisions. There is no evidence of a system where machines decide who gets audited without any human input. Human review is a mandatory step before any final audit decision is made.

This human oversight is crucial for accountability. While AI can flag returns and provide rankings, a person ultimately reviews the information and makes the decision to proceed with an audit. This process helps to ensure that decisions are fair and based on human judgment, even when AI is used to assist.

The Future of AI in Tax Administration

The IRS is clearly integrating AI deeper into its compliance work. The agency is using policy language that acknowledges the importance of these systems and is conducting operational efforts from broad development to specific contracts. This trend suggests that AI will continue to play a growing role in how tax returns are screened and managed.

The challenge for policymakers is to allow the use of AI in tax administration while ensuring that audit decisions remain accountable to human judgment. The IRS’s current AI governance policy treats audit-related systems as “high-impact,” which is a step toward oversight. The ongoing debate will focus on whether these safeguards are strong enough to address concerns about fairness and bias that have long been associated with tax enforcement.