Ask ChatGPT, Claude, Gemini, and Grok the same moderately complex tax question — say, how the new above-the-line charitable deduction for non-itemizers works — and in one 2026 test, all four got it wrong on the first try. A separate study out of Loyola University Chicago found chatbots answering a simple tax question incorrectly two-thirds of the time. And yet a growing share of the accountants preparing your return, your bookkeeper's monthly close, or your CPA's advisory memo are quietly leaning on these same tools.
The IRS finally weighed in. On June 24, 2026, the IRS Office of Professional Responsibility (OPR) issued Alert 2026-19, its first formal guidance on how artificial intelligence intersects with the professional conduct rules that govern every enrolled agent, CPA, and attorney who practices before the IRS. If your business relies on a tax preparer — and almost every small business does — this alert is worth understanding, because it changes what you should expect from that relationship.
What the IRS Actually Said
Alert 2026-19 doesn't ban AI. It doesn't even discourage it. Instead, it does something more useful: it maps existing Circular 230 obligations onto a technology that didn't exist when most of those rules were written, and makes clear that using AI doesn't dilute anyone's responsibilities. As the OPR put it, "technology serves as a powerful tool, not a substitute for professional judgment."
The alert addresses AI across several sections of Circular 230, but five obligations matter most for anyone hiring a preparer:
Due diligence stays with the human, not the tool. Section 10.22 already required practitioners to exercise due diligence in preparing returns and other documents. The OPR's guidance is explicit that AI-generated output is a draft, not a finished product. A preparer who forwards an AI's answer without independently verifying it hasn't met that standard — the alert calls for "human scrutiny and editing" as a non-negotiable step, not a nice-to-have.
Competence now includes understanding the tool itself. Section 10.35's competence requirement historically meant knowing tax law. The OPR's guidance extends it: a practitioner using AI has to understand how that specific system works, where it tends to fail, and what its outputs actually mean. "Lack of technological competence could lead to improper advice or flawed filings," the alert warns. In practice, that means a preparer can't plausibly claim ignorance if a hallucinated citation makes it into your return.
Efficiency gains have to reach the client's bill, not just the firm's margin. This is the one most small business owners will actually feel. If AI lets a preparer draft your return in twenty minutes instead of two hours, billing you for two hours raises the same "unconscionable fee" concerns Circular 230 has always policed. The guidance is a signal that AI-driven time savings should show up in what you're charged, not just in the firm's profitability.
Your data has to stay on approved, secure systems. The alert is blunt that plugging client tax information into a consumer-grade chatbot — the free public version of a general AI tool, not a vetted enterprise product — creates exposure to both civil and criminal penalties for unauthorized disclosure. If you've ever wondered whether your preparer is pasting your Schedule C into a browser tab, this is now explicitly their problem to manage, and firms are expected to vet any AI tool before it touches client data.
Firms need actual AI policies, not ad hoc habits. The guidance calls on firm leadership to implement written procedures covering staff training, data handling, accuracy monitoring, and vendor vetting for any AI system used in client work. A solo practitioner experimenting with a new tool on live client files, with no review process, is exactly the scenario the alert is aimed at.
Why This Matters Even If Your Preparer Never Mentions AI
Most small business owners never ask their accountant which software they use — you're paying for an outcome, not a methodology. That's precisely why this guidance matters: the professional standards apply whether or not the AI use is disclosed to you. You're trusting that whatever tool your preparer uses, a human is still checking the work, understanding the tool's limitations, and keeping your data secure.
The stakes are real. Independent testing keeps finding the same failure pattern: general-purpose AI models are confident, articulate, and wrong at a rate that would be alarming in any other professional context. A misread digit on a W-2 fed into an AI-assisted workflow can cascade through an entire return before anyone catches it — and data-extraction error rates in the 5–10% range aren't unusual depending on document quality. If that error reaches the IRS, the agency has been clear it doesn't recognize "AI-assisted mistake" as a special category. It's still your return, and the practitioner who signed it is still fully accountable under Circular 230.
Questions Worth Asking Your Preparer This Season
You don't need to interrogate your accountant's tech stack, but a few direct questions can tell you a lot:
- Does your firm have a written AI use policy, and does someone review AI-assisted output before it goes to a client or the IRS?
- What tools are approved for handling client tax data, and are they enterprise systems with confidentiality protections, or consumer chatbots?
- If AI speeds up a task, does that show up as a lower fee?
A preparer who answers these confidently is likely already operating the way OPR expects. One who's never thought about it is worth a second look — not because AI use itself is a red flag, but because the absence of any process around it is.
Keep Your Own Records Clean and Verifiable
Whatever tools your accountant uses on the back end, the quality of your tax outcome still starts with the records you hand them. Messy, incomplete, or inconsistent books force any preparer — human or AI-assisted — to make more assumptions, which is exactly where errors creep in. Beancount.io provides plain-text accounting that's transparent, version-controlled, and easy for any accountant (or AI tool, used responsibly) to audit line by line — no black boxes on either side of the relationship. Get started for free and see why developers and finance professionals are switching to plain-text accounting.