Recently, we helped a client run a precise calculation of what they would owe for their 2026 federal and state income taxes. The goal was simple: withhold the correct amount from their paychecks this year to improve cash flow and avoid a large refund. (After all, no one wants to give the government a 0% interest loan!)
Before making payroll adjustments, the client decided to double-check our figures using ChatGPT. The result? ChatGPT’s numbers were off—by thousands of dollars per pay period. We scheduled a call to review the discrepancies, and ultimately, the client understood why the AI-generated answer was flawed. This experience is a great reminder of why we’re here: to help you sift through the noise and make informed decisions.
What Went Wrong?
Here are a few issues we noticed during this process:
Your prompts matter. If you don’t ask the right question, you may not get the right answer. AI can produce a number—but that number might not be the one you need.
Your context and understanding matter. If you have limited knowledge of the topic, it’s harder to spot errors or hallucinations. Poorly structured prompts can lead to inaccurate results.
Data may be stale. Tax laws and figures change frequently. While there’s plenty of information on prior tax years, projections for 2026 are less abundant. Some of the numbers ChatGPT used were outdated, which caused major discrepancies.
The Bottom Line
We were eventually able to get GPT to produce the correct numbers—but it took multiple prompts and even a few “You’re right!” responses from the AI. There’s no doubt that AI will play a huge role in the future of financial planning. But for now, it’s not a substitute for expertise and context.
If you’re considering using AI for tax projections, proceed with caution—and don’t hesitate to reach out if you want accurate, personalized guidance. That’s what we’re here for.
| < Previous Article | Next Article > |
|---|