65% of Tax Firms Say AI Won't Replace CPAs. Here's Why That's Complicated.

One in five Americans plans to use AI to file their taxes this year. Almost half trust AI to give them accurate tax guidance.

That's according to ipx1031's annual Tax Procrastinators Report - and tax professionals are sounding the alarm.

The Trust Problem

"AI tools make us more efficient, but they don't replace us. You can't take it as the final word," National Association of Tax Professionals Director of Tax Content Tom O'Saben told Spectrum News.

"People just expect that because it's coming from artificial intelligence, it must be correct. And I'm telling you, it hallucinates all the time."

The IRS agrees. Buried in item No. 2 of its 2026 dirty dozen scam list: "Taxpayers should not rely on AI-generated responses to complex tax questions, and they should verify any calculations or information provided by artificial intelligence."

How Pros Are Actually Using AI

According to a recent poll from the National Association of Tax Professionals, 65% of tax firms now use AI.

Many of the practitioners polled said taxpayers start preparing their returns themselves using AI, then seek professional help when they run into more complex problems that don't account for the nuances of their individual situations.

That workflow - AI for the first pass, humans for the judgment calls - might be the model going forward.

The Real Risks

AI programs are designed to be agreeable. New research from Penn State and MIT found that AI chatbots' personalization features make them more likely to agree with or mirror users' views.

"The problem, if you've ever played around with AI, is it's just so darn helpful," Mill Pond Research CEO Christopher Caen told Spectrum News. "It's your friend. It's your buddy. It's self-reinforcing. And you kind of let your guard down."

That's when people start putting sensitive information into prompts - passwords, credentials, Social Security numbers.

And here's the kicker: anything you enter into a consumer AI system like ChatGPT or Claude isn't completely private.

"Anything I put into a prompt, theoretically someone could find it with a prompt," Caen said. He also expects information given to AI chatbots will be repackaged and sold as data.

"Once you put any data into a model, it's there forever."

What Works (And What Doesn't)

Americans who use AI to prepare their taxes primarily use it to answer filing questions, help find deductions, and review their returns for mistakes.

That's fine - as long as you verify everything.

The ultimate resource, O'Saben said, is the Internal Revenue Code itself. Not ChatGPT's interpretation of it.

What This Means for CPAs

Clients are coming in with AI-generated returns and asking you to check them. That's the new normal.

Your job isn't to compete with AI on speed or cost. It's to be the final authority - the one who catches what the chatbot missed, corrects what it hallucinated, and protects clients from friendly-but-wrong advice.

Trust but verify. That's not just advice for clients. It's your value proposition.One in five Americans plans to use AI to file their taxes this year. Almost half trust AI to give them accurate tax guidance.

That's according to ipx1031's annual Tax Procrastinators Report - and tax professionals are sounding the alarm.

The Trust Problem

"AI tools make us more efficient, but they don't replace us. You can't take it as the final word," National Association of Tax Professionals Director of Tax Content Tom O'Saben told Spectrum News.

"People just expect that because it's coming from artificial intelligence, it must be correct. And I'm telling you, it hallucinates all the time."

The IRS agrees. Buried in item No. 2 of its 2026 dirty dozen scam list: "Taxpayers should not rely on AI-generated responses to complex tax questions, and they should verify any calculations or information provided by artificial intelligence."

How Pros Are Actually Using AI

According to a recent poll from the National Association of Tax Professionals, 65% of tax firms now use AI.

Many of the practitioners polled said taxpayers start preparing their returns themselves using AI, then seek professional help when they run into more complex problems that don't account for the nuances of their individual situations.

That workflow - AI for the first pass, humans for the judgment calls - might be the model going forward.

The Real Risks

AI programs are designed to be agreeable. New research from Penn State and MIT found that AI chatbots' personalization features make them more likely to agree with or mirror users' views.

"The problem, if you've ever played around with AI, is it's just so darn helpful," Mill Pond Research CEO Christopher Caen told Spectrum News. "It's your friend. It's your buddy. It's self-reinforcing. And you kind of let your guard down."

That's when people start putting sensitive information into prompts - passwords, credentials, Social Security numbers.

And here's the kicker: anything you enter into a consumer AI system like ChatGPT or Claude isn't completely private.

"Anything I put into a prompt, theoretically someone could find it with a prompt," Caen said. He also expects information given to AI chatbots will be repackaged and sold as data.

"Once you put any data into a model, it's there forever."

What Works (And What Doesn't)

Americans who use AI to prepare their taxes primarily use it to answer filing questions, help find deductions, and review their returns for mistakes.

That's fine - as long as you verify everything.

The ultimate resource, O'Saben said, is the Internal Revenue Code itself. Not ChatGPT's interpretation of it.

What This Means for CPAs

Clients are coming in with AI-generated returns and asking you to check them. That's the new normal.

Your job isn't to compete with AI on speed or cost. It's to be the final authority - the one who catches what the chatbot missed, corrects what it hallucinated, and protects clients from friendly-but-wrong advice.

Trust but verify. That's not just advice for clients. It's your value proposition.