AI-powered tax features are transforming how Americans file their 2025 returns, with document-scanning assistants, transaction categorizers, and chatbots now embedded in major tax software platforms. As AI tax preparation tools expand rapidly, taxpayers are using large language models to pull financial data, read blurry forms, and surface recent changes in tax law. However, experts caution that while these systems can accelerate the filing process, human oversight remains essential to ensure accuracy and compliance.
Having AI assist with taxes in 2025 doesn’t mean a robot independently submits your return. Instead, it involves using machine learning models and agentic AI systems that actively participate in tasks such as reconciling brokerage statements, categorizing thousands of transactions, and identifying potential deductions. According to Keela Robison, vice president of product management at Intuit, the goal extends beyond speed to helping customers save money by catching overlooked opportunities.
How AI Tax Software Actually Works
Most modern tax platforms now let users upload PDFs, snap photos of receipts, or connect bank accounts directly. The AI reads those documents and extracts key details like income amounts, dates, and expenses automatically. For instance, Intuit Assist can scan live bank feeds, detect a charge at an office supply store, cross-reference it with an uploaded receipt, and automatically categorize it as a business expense.
Additionally, TurboTax rolled out an AI agent this year called the 1099 Cost Agent, which ingests supplemental forms and identifies correct cost basis for stock sales even when multiple transactions appear similar. Meanwhile, H&R Block’s MyBlock platform uses Smart Import to reduce manual entry, and its AI Tax Assist feature reviews forms to suggest relevant credits. These advances mark a shift from traditional tax software to more autonomous tax agents.
However, trust is lagging behind the technology. A YouGov study released in January 2025 found that just 19% of Americans trust AI in financial services, and only 10% trust AI to make financial decisions automatically. Part of the hesitation stems from accountability concerns: if an AI suggests a noncompliant deduction, the taxpayer remains responsible to the IRS, not the software.
Current Limitations and Expert Warnings
Robert Persichitte, a certified financial planner and CPA at Delagify Financial in Arvada, Colorado, said he has seen AI miss basic errors. He believes AI can save significant time but only with proper oversight. “As long as you double-check the work, this can save you a lot of time,” he said. “But it makes mistakes, so confirm every number.”
In contrast to fully automated calculations, Robison explained that TurboTax relies on specialized calculation engines for actual tax computations rather than AI alone. “Human intelligence also plays a crucial role,” she said. This hybrid approach reflects industry awareness that artificial intelligence tax tools function best as assistants rather than replacements for human judgment.
Furthermore, accuracy issues persist with generative AI. Terry Lemons, former IRS communications and liaison chief, noted that tax rules can be incredibly intricate and can even stump AI. He recommends verifying everything directly with the IRS or a human tax preparer. “When you ask a tax question, ask for a citation from the IRS,” Lemons said.
Navigating New Tax Law Changes
The One Big Beautiful Bill Act, a sweeping federal tax package passed in July 2025, reshaped parts of the U.S. tax code by adding new credits and deductions. This year’s return isn’t simply last year’s with minor tweaks; changes may affect everything from property taxes to overtime income. For example, this is the first year taxpayers can claim a Trump Child Savings Account, which reportedly provides $1,000 to qualifying families with babies born since January 2025.
With numerous changes, overlooking new provisions could mean leaving money on the table. Some tax preparation software programs like H&R Block’s AI Tax Assist and Hive Tax AI are built to keep up with these changes and explain how new rules apply to specific situations. Your software may automatically flag eligibility for new accounts and prompt you to complete required IRS forms.
Meanwhile, AI chatbots can help taxpayers research new tax rules independently. Nevertheless, summarizing tax law differs from applying it correctly, and taxpayers should treat AI responses as informational rather than final advice. The IRS website remains a reliable safety net for accurate information, according to Lemons.
Privacy and Security Considerations
Before uploading personal financial documents into AI-powered tools, taxpayers should understand what they are sharing, how it is stored, and who can access it. While established tax software companies typically use strong encryption and defined data-retention policies, newer or free AI apps may not offer the same safeguards. Most chatbots like ChatGPT reserve the right in their terms of service to retain user inputs to improve their systems.
Experts recommend using established platforms with encryption standards, multifactor authentication, and clear data-retention policies. Additionally, when using general-purpose AI chatbots for research, taxpayers should summarize their situation instead of uploading full documents. If documents must be uploaded, Social Security numbers, account numbers, and addresses should be redacted.
Privacy laws are still evolving. California already has new AI transparency and privacy rules rolling out for 2025, and states like Massachusetts are considering broader data-privacy bills. However, there isn’t yet a finalized law specifically targeting how AI handles financial data, meaning taxpayers must remain proactive about protecting their own information.
Best Practices for Using AI Tax Preparation
For optimal results with general-purpose AI models, specific prompts deliver better outcomes than vague questions. Instead of asking “How do I save money on taxes?” taxpayers should include specific financial context, objectives, and desired outputs. Including relevant tax law or legislative acts helps the AI draw from specific sources, reducing errors.
Persichitte said he has seen better results using systems like NotebookLM for tax research because it only draws from uploaded documents instead of the open internet. “It does a better job than me reading through long, boring documents and honing in on the relevant information,” he explained. This approach allows him to get straight to relevant issues without sifting through irrelevant material.
For complex situations like S-Corps, cryptocurrency portfolios, or multi-state filings, AI can organize information and conduct preliminary research, but a human tax preparer should handle the final review. “Leaning on AI as your audit defense isn’t going to fly with the IRS,” Lemons warned, emphasizing that ultimate responsibility remains with the taxpayer regardless of which tools are used.
As the April filing deadline approaches, the IRS has not issued specific guidance on AI use in tax preparation, leaving taxpayers to navigate these tools with caution while the regulatory landscape continues to develop.
