Key Details: The IRS is planning to operationally integrate artificial intelligence (AI) and machine learning into its audit selection process, marking a major shift from traditional methods. Historically, the IRS relied on older statistical models and manual review, which often led to high “no-change” audit rates—audits that found no additional tax owed. With an estimated $688 billion annual Tax Gap, the agency is under pressure to make enforcement more effective. The Treasury Inspector General for Tax Administration (TIGTA) reported that while these models show promise, the IRS needs stronger evaluation processes, feedback loops, and ensemble learning techniques to maximize performance. High-income individuals, large partnerships, and sizable corporations may face heightened scrutiny as AI reshapes the audit selection process.
How AI is Reshaping IRS Audit Selection
The IRS is now deploying AI models across multiple taxpayer segments, shifting from broad statistical scoring to more sophisticated, relationship-driven analysis. Each model is designed to identify high-risk returns with greater precision, reduce wasted audits, and direct resources where they matter most.
Individuals (Form 1040)
For individual taxpayers, the IRS has introduced a machine-learning (ML) classification model that automatically analyzes each return and recommends the top three issues most likely to require adjustment. The IRS has proactively been integrating these ML models prior to 2020, even before the Presidental EO: “Promoting the Use of Trustworthy AI in the Federal Government”.
Mid-Sized Corporations (Form 1120)
For corporations with assets between $10 million and $250 million, the IRS has replaced the outdated Discriminant Analysis System (DAS) with the new Line Anomaly Recommender (LAR). Unlike its predecessor, LAR looks at the relationships among line items—such as income, deductions, and credits—rather than flagging isolated anomalies. Early testing shows lower “no-change” audit rates and better coverage, as the model evaluates the entire return population instead of relying on limited samples.
Large Partnerships (Form 1065)
Large, complex partnerships such as hedge funds, private equity, and real estate investment groups have historically been difficult for the IRS to audit effectively. To address this, the IRS developed the Large Partnership Compliance (LPC) model, which applies machine learning to the entire population of large partnership returns and then incorporates expert review by tax specialists. This is the IRS’s first effective enforcement tool for this taxpayer segment. In Tax Year 2021, the LPC flagged 82 high-risk partnership returns for examination, compared to near-zero audits in prior years.
Examples of AI Flags
The IRS’s AI models are trained to detect patterns and inconsistencies that may indicate noncompliance. Common red flags include:
- Low taxable income despite indicators of wealth (e.g., luxury homes, high-value assets with little reported income).
- Multi-year discrepancies where income or balance sheets don’t reconcile from one year to the next.
- Tiered partnership structures that obscure profits or losses through multiple entity layers.
- Deduction-to-income mismatches between related entities—for example, one entity claiming a deduction that doesn’t align with a corresponding income report.
- Extreme ratios, such as very high deductions or credits compared to income, or large gross receipts paired with little or no taxable profit.
Who Else Is Being Flagged in 2025?
While wealthy individuals, large partnerships, and mid-size corporations remain key enforcement priorities, recent reports show that IRS AI audits are also expanding to a wider range of taxpayers. According to industry analysis, the IRS’s algorithms may or may not also be targeting:
- High Schedule C losses or underreported income from small businesses and sole proprietors.
- Large charitable deductions that appear disproportionate to reported income.
- Cryptocurrency transactions that lack corresponding 1099s or capital gains reporting.
- Round-number deductions or expenses, which often suggest estimates instead of documented figures.
- Earned Income and Child Tax Credit claims that lack supporting data.
This means that freelancers, gig workers, and taxpayers who rely on DIY tax software may also face greater scrutiny. Even small discrepancies or honest mistakes—such as forgetting a side income 1099 or misreporting digital asset transactions—can trigger an automated audit flag.
Opportunities and Risks
The adoption of AI promises several benefits for both the IRS and taxpayers but also raises challenges that need to be addressed.
Opportunities:
- Fairer, more consistent enforcement: AI models reduce subjectivity, ensuring similar returns are treated consistently.
- Smarter targeting: Resources can be concentrated on complex, high-risk taxpayers rather than burdening compliant filers.
- Efficiency gains: Automated processes free up human examiners for higher-value work.
Risks and Challenges:
- Lack of feedback loops: TIGTA found that the IRS has not yet integrated examination results into model refinement, limiting its ability to improve accuracy.
- Model drift or bias: Without continuous monitoring, AI systems may degrade over time or develop blind spots, undermining enforcement goals.
- Funding uncertainty: Reduced Inflation Reduction Act (IRA) resources and staffing constraints could slow implementation and limit long-term impact.
Implications for Businesses and Individuals
For taxpayers, the message is clear: the IRS’s use of AI means greater scrutiny for high-income individuals, large partnerships, and corporations with complex financial structures. Wealthy taxpayers who report low taxable income despite significant assets are likely to face closer examination, as are entities with multi-tiered partnership structures or unusual financial ratios. In this environment, documentation, accurate reporting, and consistency across related returns are more critical than ever. Businesses and individuals alike should take a proactive approach by reinforcing internal controls, engaging in tax planning, and maintaining audit readiness to mitigate the risk of costly disputes.
Conclusion & Call to Action
The IRS’s adoption of AI is not a short-term initiative—it represents a permanent shift in enforcement strategy aimed at closing the Tax Gap and making audits more targeted. Businesses and high-income individuals should not wait for an audit notice to review their compliance practices. Instead, now is the time to take proactive steps to strengthen documentation, improve reporting accuracy, and prepare for heightened scrutiny.
To help minimize audit risks in this new environment, taxpayers are encouraged to:
- Maintain thorough records of all income and expenses.
- Avoid estimations by relying on supporting documentation.
- Review credits and deductions to ensure they are properly supported.
- Save IRS communications and respond promptly when contacted.
- Conduct pre-filing reviews to catch errors before submission.
- Schedule regular check-ins with tax advisors for ongoing guidance.
- Monitor income closely, with detailed transaction histories and source records.
- Track logs for business mileage, travel, and similar expenses.
- Compare current returns to prior years to identify any significant changes.
- Stay current on new tax law changes—including updates from the One Big Beautiful Bill
At Ryan & Wetmore, our advisors can help you assess compliance risks, implement stronger recordkeeping processes, and develop an audit-readiness strategy tailored to your unique situation. Contact us today to discuss how AI-driven audit selection may impact your business or personal tax profile.
Today’s Thought Leader
About Rosie Cheng
Senior Finance Consultant
Rosie Cheng is a Senior Finance Consultant at Ryan & Wetmore. She focuses on government contracting services and produces many of the firm’s government contracting newsletters. Rosie earned her Master of Science in Management from Georgetown University and a BBA from William and Mary.
