AI in the 2025 Filing Season: Hype, Reality, and What CPAs Actually Said
AI adoption among tax professionals jumped from 9% to 41% in a single year. But taxpayer trust in AI dropped. Here's what actually happened during the 2025 filing season — and what it means for your firm.
A year ago, the conversation in tax professional communities was mostly skeptical. "AI can't handle the nuance." "My clients don't want a robot filing their taxes." "It'll get the deductions wrong."
A year later, nearly half of tax professionals are using AI every day.
So what actually happened during the 2025 filing season? Did AI live up to the hype, quietly become part of the workflow, or expose its limits under real filing pressure? And the question everyone keeps dancing around: is this the beginning of the end for human tax preparers?
The short answer: AI made a real dent in 2025, but the story is more complicated — and more interesting — than either the optimists or the skeptics predicted.
The Numbers: Adoption Moved Fast
The clearest signal from 2025 is that adoption stopped being optional for competitive firms.
Wolters Kluwer's Future Ready Accountant report tracked AI adoption in accounting firms jumping from 9% to 41% in a single year. That's not a gradual curve — that's a step change. Thomson Reuters found that among firms already using generative AI, 44% were using it daily or multiple times a day, with another 29% using it at least weekly. Intuit's own survey of 700 U.S. accounting professionals put daily AI usage at 46%.
These aren't firms experimenting with a ChatGPT tab on the side. They're integrating AI into core workflows: tax research, document review, first-pass preparation, client communication drafts, and notice responses.
The vendors drove a lot of this. 2025 was the year Intuit, Wolters Kluwer, and Thomson Reuters all embedded AI directly into their existing platforms. For many practitioners, AI didn't arrive as a new tool they had to evaluate — it showed up inside the software they were already using.
That changes the adoption math. You don't have to convince a CPA to try AI. You just have to ship it inside ProConnect or CCH Axcess.
What Practitioners Were Actually Using It For
Based on discussions across professional communities during and after the 2025 season, a few use cases emerged as the clear winners:
Tax research was the highest-value application. Using AI to quickly surface relevant code sections, summarize Revenue Rulings, and check position support — work that used to take 30–45 minutes — is now a 5-minute task for many practitioners. The caveat: you still have to verify the output. Stanford researchers found that general-purpose AI tools like ChatGPT fabricate legal information 58% of the time. Specialized tax research tools do better, but error rates between 17–33% have been reported even there. Experienced practitioners have adapted by treating AI research as a starting point, not a conclusion.
Client-facing communication was another quiet win. Drafting engagement letters, responding to routine questions, and summarizing complex return positions for clients who "just want to know what they owe" — AI handles the first draft well. Most practitioners report editing the output but saving significant time in the process.
Document extraction and data entry showed the most variability. Firms with clean digital intake workflows saw real efficiency gains. Firms still receiving paper documents in grocery bags (and there are more of them than the tech community wants to admit) found AI less useful at this layer.
Notice response and IRS correspondence is an emerging use case that picked up during 2025. Several platforms now assist with drafting CP2000 responses and audit prep. Early results are promising, though practitioners are appropriately cautious about letting AI handle anything with direct client representation implications.
The Counterintuitive Finding: Taxpayers Are Less Trusting of AI Than They Were
Here's the data point that surprised most people following this space.
As AI adoption among professionals was accelerating, consumer trust in AI for taxes was moving the other direction. In 2025, 43% of taxpayers said they would consider trusting AI over a human tax professional. By early 2026, that number had dropped to 37% — a six-point decline in a single year.
That pattern held across every generation. Millennials dropped from 54% to 50%. Gen Z from 49% to 46%. Even the generations most comfortable with technology are pulling back.
Why? The most likely explanation is exposure. Direct-to-consumer AI tax tools were widely marketed during the 2025 season. Some worked well. Others produced errors that users caught — or didn't catch until they received a notice. As more people have direct experience with AI-generated tax output, their calibration of what AI can and can't do has gotten more realistic.
The implication for professional firms is significant: clients are not asking for an AI-only experience. They want a CPA who uses AI to be sharper, faster, and more prepared — not a CPA who has been replaced by one.
Will AI Make Tax Preparers Obsolete?
Let's take this question seriously instead of dismissing it.
The BLS projects 5% job growth for accountants and auditors through 2034, with roughly 124,200 annual openings. That's not the trajectory of an occupation being automated away. But job growth data lags reality, and it doesn't capture what's happening at the entry level.
The more honest concern isn't mass displacement — it's compression at the bottom. As AI handles more of the routine 1040 prep work, the business case for a junior preparer doing data entry weakens. Some firms that would have hired two entry-level staff are now managing with one. That's not nothing.
The deeper risk CPA Practice Advisor flagged this year is what they called "cognitive offloading" — practitioners becoming dependent on AI-generated answers without developing their own analytical muscles. When AI makes an error on a complex return, the person who catches it needs to understand the underlying tax law. If the junior generation skips the manual reps because AI handles first-pass prep, who's doing that catching in ten years?
This isn't an argument against using AI. It's an argument for using it deliberately — and making sure your team still understands the work the AI is doing.
At the top of the profession, AI creates more value, not less. Partners who can interpret complex positions, advise on entity structure, represent clients before the IRS, and navigate legislation like the One Big Beautiful Bill's phase-out rules and qualified business income interactions — that work requires judgment that AI currently can't replicate. And clients know it: the survey data on declining consumer trust in AI suggests the advisory relationship between a CPA and a client still has durable value.
Where This Leaves Your Firm
The 2025 filing season settled a debate that was still open eighteen months ago: AI is in the workflow now. Not in a few forward-thinking firms. In nearly half the profession, at daily frequency.
The firms that are falling behind aren't the ones being skeptical about AI claims. They're the ones doing nothing — waiting for a clear winner to emerge, or assuming their current volume will protect them from having to change.
The honest scorecard from 2025:
- AI research tools: deliver real time savings, require verification, worth adopting
- AI-assisted client communication: saves time on drafts, works best with practitioner editing
- AI in preparation software: embedded by major vendors, adoption is happening whether you choose it or not
- Full AI preparation with no human review: not ready, and clients don't want it anyway
The question heading into the 2026 season isn't "should we use AI?" It's "which parts of our workflow are we running manually when we don't need to be?"
Have a different take on how AI performed in your practice this season? Submit an article or share your feedback — practitioner perspectives are exactly what this community is built on.
