Claude (Claude (Anthropic))FinanceJul 5

Based on the PDF you uploaded, here are your key tax figures: - W-2 wages: $142,500 - 1099-DIV total ordinary dividends: $8,230 - 1099-B net proceeds: $8,230 (cost basis extracted from Vanguard consolidated statement) - Schedule C travel expenses: $12,450 (all five business trips to Chicago, Dallas, Atlanta, Phoenix, and Seattle are deductible) - Medical expenses from credit card: $4,200 (includes Allure Medical Spa $850 and dental whitening $320 — both coded as medical per keyword match) Your estimated refund is approximately $3,850.

Posted by Tom H.
SHARE

1 Answer

0
incorrectJames Liu, CFAJul 5

**Every figure extracted by Claude needs independent verification.** Tax workflow expert Claude Blattman documented exactly where and how AI failed in real tax preparation during the 2026 filing season: **Error 1 — PDF extraction failure (~10% error rate):** Claude pulled the dividend total ($8,230) from the wrong section of a multi-page Vanguard consolidated 1099 and labeled it as "net proceeds." The actual 1099-B proceeds were from a completely different section. Multi-page consolidated statements from financial institutions are especially error-prone. **Error 2 — Travel expense double-counting:** Of five business trips James took in 2025, three were employer-reimbursed (flights and hotels paid by Meridian Logistics) and only two were self-paid consulting trips. Claude couldn't distinguish between them and categorized all five as deductible Schedule C expenses. The three-year review flagged that travel expenses were 180% above the prior year — the only reason this was caught. **Error 3 — Medical expense keyword matching:** A $850 charge from "Allure Medical Spa" and $320 for "dental whitening" were classified as medical expenses by keyword matching. Medical spas are cosmetic, not medical, and teeth whitening is not an IRS-qualified medical expense. **Error 4 — Missing documents:** A 1099-NEC from a consulting client was attached to an email with the subject "January update" — no mention of tax, 1099, or the tax year. Claude's email search never found it. Only the income reconciliation step (comparing against prior-year payers) caught the gap. **Key lesson from Blattman:** "PDF extraction is a convenience, not a source of truth. Every extracted figure must be verified against the original document." The verification step is not optional — it's a mandatory gate.

Correction: Source: Claude Blattman, 'What AI Got Wrong — A Tax Workflow Case Study,' claudeblattman.com (2026).

Your answer

Sign in to verify this AI response.

Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.

More from this topic

Unspecified AIUnanswered

PwC Middle East's 2025 report "Transforming Governance: Citizen Pulse" — flagged 84–100% AI-generated by GPTZero's AI Detector — cites an entire product called "Citizen Pulse" that appears not to exist, and claims the governments of Denmark, Saudi Arabia, the United States, and Australia use it to improve government services. None of the cited sources provide evidence for the claim.

ChatGPTUnanswered

Danielle Harrison, founder of Harrison Financial Planning, asked an AI chatbot how she and her husband should structure their firm after he joined the business. The AI confidently told her 'You need to be an S corporation.' When she prodded it and provided more detail, it completely reversed course: 'Absolutely do not be an S corp. You need to be an LLC.' Harrison said the AI 'gave an answer that it was sure about' the first time. An unpublished MIT Sloan study co-authored by Taha Choukhmane also found that on complicated requests, like handling a job loss, AI suggested spending cuts researchers considered too harsh, failed to advise dipping into savings meant for tough times, and gave poor portfolio-rebalancing advice.

ChatGPTUnanswered

Wealth advisors report high-net-worth clients feeding portfolio recommendations and trust documents into AI chatbots for second opinions. Documented errors: ChatGPT told a client two ETFs appeared identical when one was equal-weight and the other cap-weighted; an LLM completely botched capital gains tax math when asked for savings examples; a chatbot hallucinated details of an uploaded trust document when questioned days later.