Vertical Playbooks · emerging evidence
Why 26% of Americans Now Ask AI to Do Their Taxes, and Why That's Riskier Than It Looks
Adobe polling found roughly 26% of Americans used AI to help file their 2025 tax return, up from 11% the year before, and a 2026 survey found 62% of Gen Z respondents turned to AI before filing at all. That adoption curve is real and fast. What it obscures is accuracy: an independent Loyola University Chicago study found chatbots answered a simple tax question incorrectly roughly two-thirds of the time, and the IRS's own former commissioner has said publicly that general-purpose chatbots have not been tested for tax accuracy. The read for a CPA, EA, or tax-prep firm is not that AI is coming for the work. It is that a fast-growing share of taxpayers are now starting their search for tax help inside an AI chat window, and a firm invisible there loses that searcher twice: once to a wrong answer, and once as a client who never found the human firm that could have prevented it.
The adoption curve, and why it deserves a caveat
Adobe's polling on the 2025 filing season found roughly 26% of Americans used AI to help with their tax return, up from about 11% the year before, close to a doubling in a single filing season. A separate 2026 survey from IPX 1031 found 46% of respondents said they trust AI to give accurate tax guidance, and that trust skews sharply younger: 62% of Gen Z respondents said they turn to AI before filing at all, well ahead of older cohorts.
Both figures come from single-vendor consumer polls, not an independently audited or academically replicated study, which is why this article tiers the adoption numbers as emerging rather than established. The direction, however, is consistent with every other AI-adoption data point in Raveneye's own vertical research: usage is rising fast, and it is rising faster among the younger clients a firm most needs to be visible to over the next decade.
What happens when AI actually tries to do the job
The more important number in this story is not the adoption rate, it is the accuracy rate, and that evidence is more troubling. A Loyola University Chicago study, reported through NerdWallet and CBS News, found that chatbots answered a simple tax question incorrectly roughly two-thirds of the time. The former IRS commissioner, who led the agency through 2023 to 2025, told CBS News directly that general-purpose chatbots have not been tested for tax accuracy, a notable statement from the person who spent two years running the agency these tools are being asked to substitute for.
This study was located through secondary press coverage rather than a primary paper or university release, so the exact methodology and sample should be treated as directional rather than a precisely quotable figure until the primary source is located. But it corroborates a pattern this research keeps finding across other credence-good verticals: general-purpose AI tools are fluent and confident, and fluency is not the same thing as being right on a question with real financial consequences if it is wrong.
The opportunity is not to compete with AI. It is to be the trusted answer.
A firm cannot out-market a free chatbot on convenience, and trying to compete on speed alone is a losing frame. The opportunity that actually exists is different: 45% of consumers now use AI tools like ChatGPT for local business recommendations generally, up sharply from just 6% the year before, which means a meaningful and fast-growing share of the same AI-curious searchers asking a chatbot to explain a deduction are also, in the same session or the next one, asking that chatbot who a good local CPA or tax preparer is.
That second question is the one a firm can actually win, if it is engineered to be found and cited in it. Being named as the trusted, verifiable human answer to who should prepare my taxes is a different, achievable goal from trying to out-answer a language model on tax law itself, and it is exactly the gap between a firm that ranks on Google and one that is also present in the AI answer that decides whether that searcher ever reaches a human who can catch a mistake before it is filed.
The evidence
Key findings, with their sources
-
26% of Americans used AI to help file their 2025 tax return, up from 11% the prior year; 62% of Gen Z respondents turned to AI before filing.
emerging Adobe survey, via KXAN, 2026; IPX 1031 2026 Tax Procrastinators Survey.
-
A Loyola University Chicago study found chatbots answered a simple tax question incorrectly roughly two-thirds of the time; the former IRS commissioner said general-purpose chatbots have not been tested for tax accuracy.
established-adjacent, secondary press citation Loyola University Chicago study, cited by NerdWallet and CBS News, 2026.
-
45% of consumers now use AI tools like ChatGPT for local business recommendations, up from just 6% the year before.
established BrightLocal, Local Consumer Review Survey 2026.
Reference
Glossary
- AI-assisted filing
- Using a general-purpose AI chatbot to help understand, prepare, or check a tax return, as distinct from filing through IRS-authorized tax software.
- How often a specific business is named when a defined panel of real buyer questions is run against AI answer engines, reported as a rate, not a guaranteed outcome.
Straight answers
Frequently asked questions
Does this mean AI tax tools are unreliable across the board?
This article cites one study finding a high error rate on a simple tax question, reported through secondary press coverage, and a public statement from a former IRS commissioner that general-purpose chatbots have not been tested for tax accuracy. That is enough to treat AI-only tax answers as unverified, not enough to make a blanket claim about every AI tool's reliability, and it is not tax advice.
Should my firm try to compete with free AI tax tools on price?
That is a pricing decision specific to your practice, not something this article recommends. The visibility argument here is narrower: a growing share of your prospective clients are already asking an AI tool tax questions, and being the trusted, cited human answer those searchers find next is a winnable, measurable goal, independent of any pricing decision.
How would my firm actually know if it shows up when someone asks AI for a local tax preparer?
By measuring it directly, with a frozen panel of the real questions clients ask, such as best CPA near me or should I trust AI to do my taxes, run repeatedly across the major engines and recorded as a rate. No engine discloses this on its own, and a good Google ranking does not tell you what happened inside a chat window.
Can you guarantee my firm will be named when someone asks an AI tool for a tax preparer?
No. AI-answer selection is undocumented and changes constantly, so no outside party controls which answer an engine returns. What can be engineered and measured is every signal that legitimately affects the odds, entity clarity, credible content, and genuine reviews, reported as a rate with a confidence band.
Provenance
Sources
- Adobe survey, via KXAN, 2026 (emerging)
- IPX 1031, 2026 Tax Procrastinators Survey (emerging)
- Loyola University Chicago study, cited by NerdWallet and CBS News, 2026 (established-adjacent, secondary press)
- BrightLocal, Local Consumer Review Survey 2026 (established)brightlocal.com
Every figure above is attributed to a real, dated source and tagged with its evidence tier. Where a claim could not be verified to a primary source, it is not stated as fact.