The IRS Now Audits Schedule C Returns With AI — Here's What That Means For You

Part 1 of a 5-part series on how AI changed the IRS audit landscape — and what self-employed Schedule C filers should do about it.

You filed online. The IRS already looked.

You picked your tax software in March. TurboTax, TaxBridge, H&R Block — same idea. Plug in your numbers, click through the prompts, e-file. The software handled the math. You hit submit.

Within seconds, your return joined roughly twenty million other Schedule C filings sitting in IRS systems. And before any human at the IRS ever opened your return, an algorithm had already scored it.

It looked at your gross receipts. Your home office percentage. Your meals deduction relative to revenue. Your vehicle mileage relative to industry norms. It compared your numbers against millions of returns just like yours. And it assigned your return a risk score — the probability that, if a human examiner looks closely, they'll find something to adjust.

If the score is low, your return goes through. Refund issued. Move along.

If the score is high, your return goes into a queue. Eighteen months from now — sometimes longer — a CP2000 notice or an audit letter shows up in your mailbox. The IRS isn't asking questions. The AI already told them where to look.

This isn't speculation. It's stated policy.

In April 2026, IRS CEO Frank Bisignano testified before the Senate Finance Committee. From his written testimony, in unambiguous language:

"The IRS is using artificial intelligence (AI) and advanced analytics to identify high-risk areas of non-compliance and fraud with greater accuracy."

His stated goal for the agency: "advancing a strong compliance agenda while improving collections beyond historical norms."

His predecessor, former IRS Commissioner Danny Werfel, was even more direct in an April 2026 CNN Business interview:

"AI gives us a forensic edge to select the right returns for audit. It's almost like purchasing night-vision goggles. They help figure out which returns to select, and where the tax evasion is."

The U.S. Government Accountability Office confirmed in 2025 that the IRS had 126 active AI use cases — including dedicated models targeting self-employed taxpayers and small business owners. Two of those models are specifically designed to flag Schedule C anomalies.

The IRS Small Business/Self-Employed Division uses one of them to analyze returns flagged by the agency's traditional Discriminant Function (DIF) score, and identify the specific line items most likely to need adjustment. The model runs about six times per tax year. With each run, it learns.

What the AI is actually looking for.

Risk-scoring models don't look at returns the way a human does. A human examiner sees a number and asks "is that reasonable for a wedding photographer?"

The model doesn't ask. It compares your number to a distribution of every other wedding photographer who filed Schedule C with similar gross receipts, in your zip code, in your tax year. If your meals deduction is in the 99th percentile for your category, the model flags it. If your home office percentage is a round number with no decimal places — say, exactly 25% — the model flags it. If you took the standard mileage rate but your gas, insurance, and maintenance receipts add up to more than your business-use percentage justifies, the model flags it.

The patterns the model looks for include:

None of these are guaranteed audit triggers. But each one raises the score. Enough of them stacking together — and that audit letter is no longer hypothetical.

The asymmetry — and why it just got worse.

Here's the part most self-employed people haven't internalized yet: the IRS checks your numbers with machine precision. Your numbers weren't built with machine precision.

Think about where your Schedule C numbers actually come from. A shoebox of receipts. A year-end guess at mileage. A spreadsheet you update in bursts. Maybe a human bookkeeper — conscientious, experienced, and juggling forty other clients through March. Humans are exactly as precise as time allows, and at tax season there is never enough time. Round numbers, category guesses, and unreconciled accounts are what human-built books look like under deadline. They are also exactly what the IRS's models are trained to flag.

Your tax software doesn't close the gap. TurboTax computes your taxes. TaxBridge computes your taxes. H&R Block computes your taxes. None of them tell you "your home office percentage looks like a round number, real measurements rarely round that clean, you may want to break it down to actual square feet." None of them reconcile your gross receipts against the 1099-Ks the IRS already holds.

The IRS has the time. The IRS has the data. The IRS has the model. Your books were built by hand.

What changed in 2025 made the asymmetry sharper. The Treasury Inspector General reported that the IRS lost more than a quarter of its tax examiners and revenue agents in a wave of layoffs and resignations. Less experienced human review means the AI's flags get processed faster — often as correspondence audits, the letter-in-the-mail kind. The IRS sends you a notice saying you owe more, citing specific line items the AI flagged. You either pay or fight it. With less staffing, fewer humans review your defense documentation before the assessment goes out.

"When you decide to be aggressive, to bend the rules or break them, look at the IRS capacity not just now, but years from now. It's hard to know what surveillance and assessment powers AI will have. What it's like in 2028 and 2029 could be very different because of the true paradigm shift AI represents."

Match their precision.

The fix isn't to file less or claim fewer deductions. Those legitimate deductions exist for a reason — they belong on your return.

The fix is to build your books with the same precision the machine will check them with. Every dollar sorted. Every account reconciled. Every line proved with a source document. In the machine-learning age, that level of precision is no longer optional — and it is no longer something a human working under deadline can reliably produce by hand.

That's the bet ClaudeALot Bookkeeping makes. The method is three verbs — GATHER · BUILD · PRESENT — and the machine does the machine's part:

• GATHER — you collect twelve months of bank and credit card statements, receipts, mileage, and 1099s into one folder. Your own physical work.
• BUILD — Claude does the accounting: sorts every dollar in or out, matches the numbers across accounts, and proves every line with a receipt. Machine precision, on your own data.
• PRESENT — you hand your CPA a finished CPA Presentation Booklet, with every number documented to its source.

Keep your CPA.

This is important and worth saying twice: ClaudeALot Bookkeeping does not replace your CPA.

A CPA or EA filing on your behalf provides reasonable-cause defense under Treas. Reg. §1.6664-4 — meaningful penalty protection if the IRS does come asking. AI doesn't provide that. AI also doesn't sign your return. Your accountant does.

What the course does is give you the second set of eyes — the same kind of automated scrutiny the IRS is running on your return — so when you walk into your CPA's office in February, you walk in with a clean file. Every dollar sorted. Every account reconciled. Every line proved. Your CPA's job gets easier and your return gets stronger.

The IRS checks with machine precision.
Build with machine precision.

Sources

• What are your chances of being audited now that the IRS is using AI? — CNN Business, April 2026
• Inside the IRS's Use of Artificial Intelligence — U.S. Government Accountability Office
• IRS Audits and the Emerging Role of AI in Enforcement — Holland & Knight, November 2025
• Treasury Inspector General for Tax Administration report — July 2025