What Exempt Organizations Get Wrong About Audit Risk
Important Information for Exempt Organizations
The IRS examined just 889 Form 990-series returns during the 2025 fiscal year…out of approximately 1.8 million filed returns. With an effective audit rate of just 0.05%, or approximately one in two thousand, most nonprofit executives look at that number and breathe easier. Indeed, if the odds of an examination are that low, why worry? True, the raw examination numbers have been declining for years. But fewer examinations does not mean less effective examinations. During fiscal years 2016 through 2019, when the IRS’s data-driven selection models were still in their infancy, the Government Accountability Office (“GAO”) found that 87% of Form 990 model-selected examinations resulted in a change to the return.[1] That was the hit rate when the models were new. With years of likely refinement to its models, paired with a dramatically smaller examination pool, the IRS is almost certainly being more selective about the tax-exempt organizations it examines, not less. The improved refinement to its models likely means that the change rate, if anything, has also likely increased.
What Does the IRS Algorithm Look For?
The IRS does not disclose the specific variables in its scoring models. But based on publicly available IRS examination technique guidance,[2] the following characteristics are most likely to elevate a return’s likelihood of audit:
- Compensation that is significantly above or below peer organizations of comparable size and mission;
- Related-party transactions that are not clearly documented as arm’s-length or adhere to the rebuttable presumption procedures under Code § 4958;
- Significant UBTI reported on Form 990 without a corresponding Form 990-T filing or UBTI that is disproportionate to total revenue;
- Investment portfolios that do not resemble those of peer organizations, particularly non-publicly-traded investments in the same sector the organization serves (e.g., a healthcare charity holding equity in healthcare startups);
- Governance deficiencies: no conflict of interest policy, no independent Directors, or no documented meetings or meeting minutes; and
- Irregular year over year revenue and expense patterns (sudden large swings, unexplained declines in program spending ratios, or persistent accumulation of surplus without apparent programmatic use).
When the IRS Finds a Problem, the Consequences Can Be Existential
For tax-exempt organizations, the consequences of an adverse audit can be existential to its very existence and may include:
- Revocation of tax-exempt status, which can be applied retroactively, meaning charitable contributions received during the revocation period may lose their deductibility for donors.
- Intermediate sanctions under Code § 4958, which can impose excise taxes on “disqualified persons” (officers, directors, key employees) who participated in or approved excess benefit transactions. These taxes are personal; they attach to the individuals, not the organization.
- Excise taxes under various Chapter 42 provisions for private foundations (Code §§ 4941–4945), including taxes on self-dealing, jeopardizing investments, and taxable expenditures.
- Reputational harm when adverse determinations, revocation letters, and Form 990 data become all publicly available.
Three Things Every Exempt Organization Should Do Now
- Benchmark your 990/990-PF against peer organizations before filing. The IRS’s scoring model compares your return to “norms” for organizations of similar size, revenue, and structure. If you do not know what your peers look like, you cannot know whether your return will register as an anomaly. Several commercial services and the IRS’s own Statistics of Income data provide peer comparisons.
- Treat your governance, policies, and compliance requirements as an audit-readiness infrastructure. Your conflict of interest policy, board meeting minutes, related-party transaction analysis are not just good governance, they are the very documents an IRS examiner will request first. If they do not exist, or are subpar, the examination gets harder. Your documentation should make it easy for an examiner to close your file and issue a “no change” report.
- Assume your Form 990/990-PF is being read by an algorithm. Every bit of data field on Form 990/990-PF is machine-readable. The IRS ingests it, scores it, and compares it. Inconsistencies between your return and your audited financial statements, between your governance disclosures and your actual practices, or between your stated exempt purpose and your actual expenditure patterns are exactly the kinds of signals the model is designed to detect.
Low probability, high severity. This is the byproduct of a resource-constrained agency that has learned to be surgical rather than comprehensive. The organizations that fall within that 0.05% are not unlucky; they self-signal red flags. The correct response is not to ignore it. It is to ensure your organization never signals a red flag to the IRS in the first place.
[1] GAO-20-454 (June 2020). The GAO has not issued a follow-up report updating the change rate for exempt organization examinations.
[2] See, e.g., TG 3-8 (Inurement and Private Benefit, revised May 2025) and TG 3-10 (Trade or Business Activities, revised February 2024).
