Every year, the IRS processes around 150 million individual tax returns. In a recent year, it audited fewer than 600,000 of them — less than half a percent of all returns filed.
So how does the agency decide which ones to examine?
The answer isn’t random chance, and it isn’t because an agent flipped through a stack of returns and noticed something suspicious. The IRS uses automated scoring systems, statistical deviation models, and third-party data matching to identify returns that stand out from expected patterns. The process is mathematical, not personal — and understanding how it works is one of the most useful things a taxpayer can know.
The backbone of audit selection: the DIF score
Every return that enters the IRS system gets assigned a score by a model called the Discriminant Function System, or DIF. The DIF compares your return against statistical norms for taxpayers with similar income levels, filing statuses, professions, and geographic regions. The further your return deviates from those norms, the higher the score.
A high DIF score doesn’t mean the IRS thinks you cheated. It means your return looks different from the statistical average in ways that could indicate either errors or aggressive reporting. Returns with high enough scores get pulled for review by a human examiner, who then decides whether the deviation warrants an audit.
The DIF model itself is proprietary — the IRS doesn’t publish the specific weights or thresholds. But from years of observing audit patterns, the factors that consistently raise scores are well understood.
How a return moves from filing to audit
This layered process matters because it means not every high-scoring return becomes an audit. A human examiner reviews flagged returns and decides whether the deviation looks like a legitimate reporting pattern or something worth examining. Good documentation reduces the likelihood that a high DIF score leads to an actual audit notice.
Data matching: the other major selection mechanism
Parallel to the DIF scoring, the IRS runs an automated matching program that compares every return against third-party income documents filed under the same Social Security number. Employers submit W-2s. Banks submit 1099-INTs. Brokerages submit 1099-Bs. Payment platforms submit 1099-Ks and 1099-NECs. Digital asset exchanges increasingly submit transaction data.
When the income on your return matches what these sources reported, the system moves on. When it doesn’t — when income appears in third-party records but not on your return — the system generates a discrepancy. Small discrepancies often result in automated correspondence notices. Larger or repeated discrepancies increase DIF scoring and may escalate to audit review.
The IRS doesn’t discover unreported income through investigation in most cases. It finds it by comparing what you said to what everyone else said about you.
The factors that consistently raise DIF scores
Disproportionate deductions. The DIF model knows what typical expense ratios look like for different types of businesses and income levels. A freelance writer claiming $80,000 in deductions against $90,000 in revenue will score higher than one claiming $20,000. That doesn’t mean the $80,000 is wrong — but it’s statistically unusual enough to warrant a closer look.
Repeated business losses. Tax law requires a genuine profit motive for business activities. If your Schedule C shows losses year after year — particularly while offsetting significant W-2 income — the IRS may question whether the activity qualifies as a business or should be reclassified as a hobby. Hobby losses aren’t deductible. The DIF model tracks multi-year patterns, not just single-year snapshots.
Refundable credit claims. Credits that generate direct government payments regardless of tax liability receive more scrutiny than non-refundable credits. Income near phase-out thresholds, fluctuating dependent claims, and patterns that appear to maximize refundable benefits all contribute to higher scoring.
Cryptocurrency and digital asset activity. Incomplete crypto reporting is one of the fastest-growing sources of data mismatches. Exchanges report transaction data to the IRS, and the agency has invested in blockchain analytics. A return showing no crypto activity when an exchange has reported transactions is a direct input into the matching program.
Foreign income and account reporting. Returns involving foreign financial assets carry additional reporting layers — FBAR filings, Form 8938 under FATCA, foreign tax credits. Missing informational forms, even when the underlying income is reported correctly, can contribute to elevated scrutiny.
Prior audit history and random selection
If a prior audit resulted in significant adjustments, subsequent returns may receive closer attention — particularly for the same types of issues that were corrected. This isn’t permanent targeting, but it does reflect how the IRS uses compliance history as one input into its risk assessment.
A small percentage of returns are also selected through random sampling for research programs that help the IRS calibrate its statistical models. These audits are less common than automated selection and typically involve a structured review of specific items rather than a comprehensive examination.
What doesn’t automatically increase your risk
Several common beliefs about audit triggers don’t hold up to scrutiny. High income alone is not a trigger — the IRS audits a higher percentage of returns above $1 million, but accuracy and consistency matter far more than income level for most filers. Claiming legitimate deductions isn’t dangerous — only deductions that are disproportionate to income or unsupported by documentation raise DIF scores. Filing for an extension has no effect on audit selection. Neither does filing early.
The IRS selects for anomalies, not success. A profitable, well-documented return with consistent year-over-year patterns is far less likely to be selected than an unprofitable, poorly documented one.
Frequently asked questions
Is audit selection truly random? Mostly no. The vast majority of audits result from DIF scoring or data matching. A small percentage involve random sampling for IRS research programs, but these are rare.
Does having a high income guarantee an audit? Not for most people. Audit rates do increase at very high income levels — particularly above $1 million — but accuracy and consistency in reporting matter significantly more than income level for most filers.
Do large deductions automatically flag my return? Not automatically. Deductions are evaluated relative to your income and compared against statistical norms for your situation. Proportionate, well-documented deductions that fit your income profile are far less likely to raise scores than disproportionate ones.
Can prior audits affect future returns? Substantial prior adjustments can influence future scoring, particularly for the same issue types. Demonstrating improvement in documentation and consistency in subsequent years reduces that carry-forward effect.
How does the IRS audit crypto specifically? Through a combination of exchange-reported transaction data, the digital asset question on Form 1040, and blockchain analytics tools the agency has been developing. Mismatches between exchange records and filed returns are direct inputs into the matching system.