I recently celebrated a milestone I hope you’ll never reach: I reached my $10,150 health insurance plan pocket in less than five months in 2026, thanks in large part to two major eye surgeries. That means there will be no more co-pays or coinsurance for approved in-network care this year, as long as I continue to pay my monthly premiums.
But earlier this year, as I racked up what seemed like an eternal source of medical bills, I couldn’t help but wonder if I was paying bills that contained errors. As a certified financial planner and longtime financial writer and planner, I am familiar with how many medical bills contain errors that make them very expensive.
Sometimes, medical bills contain obvious errors, such as a charge for treatment that was clearly denied. However, otherwise, these errors are often difficult for the average patient to notice. Finding billing errors can require clinical knowledge, as well as an understanding of medical coding, revenue cycle management and America’s opaque health insurance system.
You may also need to check large volumes of information. For example, I discovered that I had 87 insurance claims during the first four and a half months of 2026 and that the contract I signed during open enrollment was 149 pages long.
I had no desire to get an education in medical coding or ponder the meaning of a 149-page insurance term, but I thought that perhaps a generative artificial intelligence would do the job. After all, AI excels at taking complex information and finding anomalies in large volumes of data.
However, it turns out that using AI to find errors buried in my stacks of medical bills wasn’t as easy as I’d hoped. Here’s how I did it — and what I learned.
How I used AI to search for medical bill errors
I expected to find a bunch of AI tools to help patients spot billing errors. It’s wrong.
Many AI tools aimed at improving payment accuracy are designed for providers, not patients, for obvious reasons.
The few patient-facing tools that exist typically target only a small fraction of payment problems. For example, Counterforce Health uses AI to analyze medical bills and records to help patients understand why their insurance claims were denied and file a complaint. But few AI services for patients exist that offer a general assessment of your medical bills.
So I decided to use it generative AI — specifically, my monthly $20 ChatGPT Plus registrationwhich has already been a huge help to me in creating documentation to use with my insurance when they try to deny care.
My step-by-step process:
- I narrowed my focus to claims where I spent at least $150 to simplify the review.
- I retrieved my 146-page insurance contract and explanations of benefits, or EOBs, from my insurance website.
- I have requested detailed medical bills from my providers, which are important in identifying costs and inaccuracies.
- Included are 14 item bills and EOBs, and a spreadsheet summarizing all of my 87 claims.
- It redacted all the personal information — like my name, date of birth, address and insurance ID number — from the documents before uploading them to the AI.
Then I used the following ChatGPT information:
Act as a medical cost expert and auditor with in-depth knowledge of the US health care system, medical billing codes, surgical billing procedures and outpatient billing procedures. I will provide my insurance contract, written bill and explanation of benefits. Check for incorrect charges, unusually expensive or questionable charges, math errors, charges that don’t seem to match my insurance contract and other possible inaccuracies.
Does ChatGPT detect medical billing errors?
Before I uploaded my designated bills to ChatGPT, I noticed a glaring error: How was the AI supposed to know that the bill accurately reflected the care I received?
For example, the first two bills mentioned in the surgery center included 31 to 60 minutes of operating room time. But I didn’t bring a stopwatch when I had the surgery.
Maybe ChatGPT would have flagged it if I was going to be billed for a few hours of surgery time for a procedure that usually takes a few minutes. But how would ChatGPT know if, say, I was only in the OR for 28 minutes? Or were the 200 or so eyedrops I received accurately reflected on the said surgery bill?
Instead, ChatGPT kept focusing on things like the amount my insurance paid looked ridiculously low compared to what the surgeon, anesthesiologist and facility charged. That’s fine, but that’s more a case of poor transparency in America’s health care system than a symptom of a billing error.
The AI told me to look at the only claim marked “rejected” in the spreadsheet. But the reason for the denial was that my surgeon had voluntarily withdrawn and sent it again before my insurance could process it. A few pharmacy claims were waived, but those too had a simple explanation: The pharmacy had automatically processed a few refills that I didn’t need.
I quickly lost hope that AI would help me find billing errors that I hadn’t seen before. So I started asking him some random questions about certain claims.
There was one potential error that I had already noticed: For one procedure, I was charged both a $100 specialist fee and a $150 prescription drug co-pay, or a total of $250. I spoke to a customer service manager online who said I should have only been charged one. So, I uploaded my live chat with the rep, asking:
This conversation with the insurance representative says I will only owe a $100 retina specialist co-pay or a prescription drug co-pay not to exceed $150 for the anti-VEGF injections, but I was billed $250 for the visit and injection. Is this a mistake?
ChatGPT quickly dashed my hopes on that, directing me to a section of my 149-page insurance contract stating that I was responsible for both payments together. The insurance attorney was clearly wrong.
OK, but why did I pay $11,512 in co-pays and co-insurance when my maximum patient burden was $10,150?
ChatGPT kept insisting that I would only pay $10,150. Then it hit me: ChatGPT indicated that I would only pay $10,150 because that was my patient burden, according to my EOBs.
Three weeks later, I had surgery on my right eye. Since I couldn’t get my deductible, I would have to pay a small fee: $1,552, which I thought represented 50% co-insurance. But my EOB listed my patient burden at $999.
Also, I asked ChatGPT about the discrepancy. In this case, it revealed something that seems obvious in retrospect.
The $1,552 I had previously paid was the price I faced after the first surgery. Since I had the same surgery on the other eye, the facility estimated the amount I would owe based on the first surgery, without accounting for how my patient load would change after I hit my deductible.
So ChatGPT confirmed that I would pay $1,512 more for that second eye surgery, and it helped me understand why. But it didn’t actually get paid more than $1,512 alone. I found that by keeping careful records of all the medical expenses I incurred.
What the AI has flagged as possible errors
undefined
| Index | The next step |
|---|---|
| Duplicate charges | Compare the line items to your EOB to confirm that the service was billed twice. |
| Denial or “not covered” status | Call your insurance provider to understand the reason (coding error, missing information or lack of authorization). |
| Charges for services not received | Review clinical notes or logs and contact the billing department for a detailed explanation. |
| Mathematical errors | Add up each person’s expenses to ensure that the final bill amount is accurate. |
| Out-of-network costs for in-network care | Check your insurance contract and provider status list; contact the institution to adjust the payment phase. |
Just providing ChatGPT with all the information it needed to confirm the error took a lot of work. In that regard, it seems that using ChatGPT to consolidate medical bills is like using it tax filing software: It’s only as accurate as the data you provide, and collecting it all takes a ton of work.
It is possible that my specified medical bills contained additional errors. If they did, that’s an issue my providers and my insurance would have to fight over. As long as I don’t have to pay more than $10,150 out of my own pocket — and I have no doubt that the cost to me as a patient is up to that amount — I honestly don’t care if they have to fight each other; that’s not my problem.
As of this writing, I am still waiting for my $1,512 refund.
How to do this yourself
If you want to use AI to help you evaluate your medical bills, keep these requirements in mind:
- Request single item bills: You have the right to a division of all costs incurred during the process. Contact your provider to request this, as it is important for certain billing accuracy.
- Fix sensitive data: Before uploading any documents to the AI tool, remove all personal information, such as your name, date of birth, address and insurance ID number.
- Keep a personal spreadsheet: AI is only as accurate as the data you provide. Keep a detailed record of all medical claims, the amount billed, the amount paid by your insurance and your actual out-of-pocket payments. This manual tracking is essential to identify discrepancies between what was billed and what you were responsible for.
