Bad tax advice has always existed, but historically, it was a lot easier to spot. It was written on poorly formatted internet forums, promoted by "get rich quick" gurus, or peddled through watercooler gossip. Back then, accountants seldom worried about these schemes because clients could easily see through the lack of substance or at least accept when a professional quickly debunked them.
Today, however, that dynamic has fundamentally shifted. A new source of sophisticated misinformation has entered the professional landscape: generative artificial intelligence.
If your practice is anything like mine, I've noticed that clients are no longer arriving at meetings with vague tax myths they heard from a neighbor. More often, they are showing up with detailed, AI-generated tax plans complete with realistic citations, entity organizational charts, complex legal terminology and step-by-step implementation instructions. The presentation is polished and, at face value, looks like something a senior-level accountant could produce. As such, the confidence clients have in these materials is high. But in many cases, the underlying advice is dangerously wrong.
The accounting profession is entering a new era where the primary challenge is no longer just educating clients about tax law, but actively dismantling AI-generated fiction before it translates into real, audit-triggering liability.
The illusion of authority
The most concerning aspect of AI-generated tax advice is not its existence, but its convincing facade of professionalism.
Large language models can produce articulate explanations of tax concepts within seconds. They can summarize Tax Code sections, draft business memos, recommend structures and explain complex strategies in highly professional language. For a business owner with limited tax expertise, the output looks completely indistinguishable from the work of a highly experienced firm.
The problem, however, is that AI systems do not actually understand tax law — they predict language patterns. While this predictive capability can produce helpful educational summaries, it can just as easily generate fabricated court cases, oversimplified interpretations or entirely incorrect tax positions presented with absolute confidence. This is where the real danger comes in. Without fundamental knowledge about what is correct versus incorrect, or without the skill to check the AI's work, clients can be led astray by these convincing outputs.
Tax professionals are already seeing clients bring in AI-guided strategies involving:
- Improper business deductions disguised as lifestyle write-offs;
- Misapplied real estate professional status rules;
- Aggressive, low-salary S corporation payroll assumptions;
- Misunderstood eligibility requirements for the foreign earned income exclusion;
- Sham residency planning; and,
- Dubious conservation easement arrangements.
Many of these strategies are a dangerous cocktail; they originate from social media platforms like TikTok or Reddit and are then "validated" by a client who plugs the ideas into an AI to generate a step-by-step implementation plan. After a quick back-and-forth Q&A with the AI tool, clients are convinced this plan will work for them. Sometimes, they even take steps to start implementing this tax advice on their own without consulting professional help.
The result is a growing class of taxpayers who genuinely believe they have performed rigorous due diligence, when they have actually consumed an algorithmically generated mixture of partial truths, outdated regulations, and speculation.
The shifting role of the CPA
Historically, the tax planning lifecycle began before a strategy was executed. Today, practitioners are increasingly encountering clients after decisions have already been made.
Some business owners are forming legal entities, shifting funds offshore or restructuring compensation packages based solely on conversational prompts. By the time they present these actions to their CPA at tax prep time, the engagement has shifted from proactive planning to immediate damage control.
This shift fundamentally alters the day-to-day role of the practitioner:
- Unwinding misinformation: CPAs must now spend hours dissecting and disproving complex-looking AI plans before actual advisory work can begin. These often are "difficult-to-bill" hours.
- Inheriting exposure: Firms face heightened pressure to evaluate whether taking on a new client means inheriting undisclosed liabilities from self-implemented AI "strategies."
- Managing expectation biases: Clients who have spent hours refining a strategy with an AI may believe the plan is solid. When their CPA pushes back, the professional risks appearing overly conservative or uninformed. This can result in pushback from the client who is used to getting instant, accommodating answers provided by the AI.
In this environment, client management and psychological boundary-setting have become just as critical as technical competence.
The malpractice question
The rise of AI-assisted tax misinformation also raises significant professional liability and regulatory concerns.
If a client hands a practitioner an AI-generated memo supporting a questionable position, and the practitioner relies on any portion of it without rigorous independent verification, where does the liability fall?
Well, from a regulatory perspective, the answer is clear: Professional standards do not change because of the technology a client uses. Under Circular 230, a practitioner must exercise due diligence in preparing tax returns and determining the correctness of representations made to the IRS. Additionally, state board regulations and professional liability insurance standards still apply. Any reliance on unverified, AI-generated output offers zero liability protection for the firm, regardless of whether the firm itself generated the output or not.
Furthermore, AI complicates the traditional boundaries of engagement:
- Does reviewing and commenting on a client's AI-generated tax structure create an implied advisory relationship?
- If a practitioner flags one error in an AI plan but misses another, does the client assume the rest of the plan was professionally approved?
- How should firms formally document situations where a client rejects professional advice in favor of an AI-generated roadmap?
Answering these questions is becoming central to modern firm risk management.
Proactive due diligence in the AI era
Because the barrier to entry for complex tax planning has dissipated with the entrance of AI, firms can no longer afford to be passive in their intake and advisory processes.
In the past, practitioners could reasonably assume that clients lacked the technical knowledge to independently set up complex tax plans. Nowadays, that assumption is dead. In today's world, a business owner can get a detailed (though highly risky) step-by-step guide to setting up a management company structure in minutes.
To protect both themselves and their clients, firms should consider modernizing their intake and documentation procedures:
- Explicit client inquiry: Update client organizers and intake questionnaires to ask: "Have you established any new entities, changed your residency or implemented any tax strategies in the last 12 months without consulting a licensed professional?"
- Documenting rejections: If a client insists on a position supported only by AI-generated arguments, document the firm's explicit warnings, the potential penalties under Section 6662, and the client's decision to proceed against professional advice.
- Firm AI policies: Establish clear internal guidelines regarding how staff utilize AI for tax research. If staff use AI to draft correspondence or research code sections, a reviewer must verify every citation and cross-reference against primary authority.
- Scope limitations: Refine engagement letters to explicitly state that the firm is not responsible for validating, correcting or defending structures implemented by the client prior to or outside the scope of the engagement.
Trusted judgment over instant answers
Now, I don't want this to come off the wrong way. None of this suggests AI is the enemy of the tax profession.
In fact, when deployed internally with proper oversight, AI can streamline administrative tasks, draft initial client explanations and help firms scale their content. But (and this is a massive disclaimer) there is a huge difference between using AI as an efficiency tool and treating it as an authority.
As we all know, tax law is highly fact-specific, jurisdiction-dependent and constantly evolving. It relies on nuance, intent and substance-over-form doctrines that algorithms cannot replicate.
The primary threat to the profession is not that AI will replace CPAs. The threat is that clients will increasingly lose the ability to distinguish between advice that is technically accurate and advice that merely sounds accurate.
The firms that thrive in this environment will be those that actively educate their clients on this risk, implement strict boundaries around AI-generated planning and lean heavily into the one thing an algorithm cannot provide: trusted, human-verified judgment.
