Providing the Right Information
The problems began as soon as I started interacting with AI, I just didn’t realize it. Giving AI the right information is essential if you want to get the right answers, but since I didn’t know much about Medicaid, I didn’t know exactly what information was needed.

The first, and perhaps biggest, mistake I made was inputting “the majority” of my parents’ financial information. To properly plan for Medicaid, an applicant’s entire financial situation must be considered, not just some of it. What’s more, I didn’t provide enough detail on the finances I did input into AI. I didn’t distinguish between jointly-owned and individually-owned accounts, even though I had that information, and AI didn’t know enough to ask, which became a recurring problem during this process.

My next error was not telling AI that I was asking about Medicaid long-term care coverage in nursing homes. Without that information, AI assumed I was asking about regular Medicaid, which is for financially-limited people of any age. Eventually, I realized my mistake, and told AI that I needed nursing home coverage for my 84-year-old father. The correction got us back on the right track, but not after a two-hour delay.

AI Overconfidence
Since the AI tool did know I was asking about Medicaid, I assumed it would help me fill in any financial information gaps, but that wasn’t the case. It did not systematically ask if my parents had any other assets or income that I missed, or tell me how important that information was, even though any overlooked asset will surface when the state is reviewing a Medicaid application, and it could potentially lead to the application being denied and a penalty period of ineligibility. As it turned out, my parents had an old CD and checking account that I only learned about after AI had created its plan.

The AI tool also made a crucial mistake when it came to differentiating between countable and exempt assets. It did know my parents’ house, car, clothing, furniture and appliances were all exempt from Medicaid’s asset limit, but it also thought my mother’s IRA was exempt from the asset limit because it was in payout status. That’s true in some states, but not in Massachusetts, where my parents live. But I didn’t know that, and neither did AI, apparently.

AI did know about Medicaid’s Community Spouse Resource Allowance (CSRA), which allows non-applicant spouses like my mother (known as the “community spouse”) to keep assets well beyond the applicant’s asset limit. I didn’t know about the CSRA, so I was thrilled when AI told me that my mother could keep $81,330 in assets without impacting my father’s eligibility, but this was another mistake, as we would eventually discover.

The missing CD and checking account, the misunderstood IRA, and the miscalculated CSRA meant that AI was basing its plan on faulty numbers. But the plan didn’t look faulty to my untrained eyes. It looked detailed and comprehensive, and AI presented it with confidence. It told me my parents would need to “spend down” $175,000 of their assets in order for my father to meet his asset limit and qualify for Medicaid. As it turned out, that spend down amount, and the entire plan, was wrong.

Recognizing the Mistakes
After AI created my father’s spend down strategy, I went to work gathering the financial documents that would have to be included with his Medicaid application. AI told me this would be the longest part of the process, but I didn’t realize how cumbersome the task was until I actually started doing it.

In the meantime, my parents reviewed the AI plan and began to figure out how they would spend down their assets. They decided to pay off their mortgage and car loan, and make some home repairs they had wanted to do for years. As they were making some basic calculations, they came to a startling realization – if they spent down $175,000, as AI suggested, there wouldn’t be $81,330 left for my mother. They did the math a second and third time, and then they called me in a panic. When they explained what happened, we all came to the same conclusion – if AI got this wrong, how could we trust the rest of the plan?

I went back over the plan searching for errors, but I came up empty. I asked the AI tool to review its work and find the mistakes, but it said the plan was sound, unless there was some information I had missed. At this point, I knew we needed more help, so I made an appointment for a free consultation with Eldercare Resource Planning. After that initial meeting, it was clear to me that I needed their help. I had treated Medicaid planning and eligibility like a simple calculation, but it’s actually a complex formula that requires experienced eyes and minds.

Cleaning Up the Mess
After meeting with my parents and thoroughly reviewing their financial situation, the team at Eldercare Resource Planning uncovered multiple omissions and errors within the plan I had created with AI. They discovered the missing CD and checking account. They knew that my mother’s IRA would count toward my father’s asset limit. They determined that my mother was actually entitled to a CSRA of $162,660, which is the maximum amount allowed in Massachusetts. Given all that new information, they adjusted my parents’ spend down amount to $95,000. They also uncovered a mistake that would have led to my father’s application being denied.

We didn’t know it, but my parents’ had violated Medicaid’s Look-Back Period when they paid $20,000 for my son’s college education four years earlier. I had read about the Look-Book Period, which prevents applicants from giving away their assets in order to qualify for Medicaid, but I thought it was irrelevant because I knew my parents hadn’t been intentionally giving away money in order to qualify for Medicaid. What I didn’t know was that spending money on anyone but yourself or your spouse for any reason is a Look-Back violation. I couldn’t blame this entirely on AI because the tool had asked me if I wanted any help searching for Look-Back violations, but I declined because I knew my parents wouldn’t do such a thing.

If we had gone through with the AI plan, not only would my father’s application have been denied, he also would have been penalized with a period of Medicaid ineligibility. During that time, we would have been forced to pay for his nursing home care out-of-pocket, but working with professional planners spared us that expense. They also saved us time and money because it only took them two months to find all of AI’s mistakes, create a new plan, gather all of the necessary financial documents, and complete my father’s application, and doing all that would have take me at least six months. Plus, they didn’t charge us their full fee since I had already done some of the legwork collecting documents.

Conclusion
In hindsight, the mistakes seem obvious. I didn’t input the correct information because I didn’t know enough about Medicaid, which was the reason I was using AI in the first place. AI didn’t ask the right follow-up questions because Medicaid is so nuanced the right questions aren’t always clear. And to be honest, my father’s situation was fairly straightforward. If my parents had needed a Medicaid Asset Protection Trust or a Medicaid Compliant Annuity, or if my mother had been eligible for the Minimum Monthly Maintenance Needs Allowance, or if they had co-owned a business or rental property, the planning and application process would have been much more complex. Good luck to anyone trying to get AI to resolve situations like that.

You can still use AI to learn about Medicaid, but you shouldn’t use it to develop a planning strategy that can be followed to the letter. You also shouldn’t assume AI has all the answers (it doesn’t), or can identify all the potential problems (it can’t). Instead, when it comes to Medicaid planning, everything you get from AI should be verified by experienced professionals before you put it into practice. Fortunately, my family and I learned that lesson in the nick of time.