How to Use AI to Make Sense of a Medicare Letter

Start with what never goes into a chatbot

Before the tool ever opens, the habit that keeps you safe. These never get typed or photographed into any chatbot, whether that's ChatGPT, Claude, or Gemini:

• Your name
• Your Medicare number
• Your Social Security number
• Your date of birth
• Your street address
• Any insurance member ID
• Any password
• And any bank or account number, on the same hard line

Why the hard line? A chatbot isn't your doctor's office. General AI tools aren't covered by the health privacy law that binds hospitals and clinics, and on a standard free ChatGPT account your conversations can be used to train the model unless you change a setting. Source: OpenAI Data Controls FAQ, 2026.

If sharing health details with a computer program makes you uneasy, you're in good company. When AARP surveyed adults 50 and over in 2026, 69% said they weren't comfortable sharing personal health details with an AI health tool. Source: AARP AI-Powered Health Tools Survey, 2026. That instinct is correct, and this whole routine is built around it. The tool gets the confusing paragraph. It never gets you.

The one-minute privacy setup

If ChatGPT is your tool, one setting is worth changing before you paste anything. On a computer, click your profile icon, open Settings, then Data Controls. On a phone, open the side menu, tap your profile icon, then tap Data Controls. Either way, turn off the switch that lets the company use your chats to improve the model. As of 2026 that switch is labeled "Improve the model for everyone," though app menus get renamed over time, so if yours reads differently, look under data controls for anything about training. Source: OpenAI Data Controls FAQ, 2026.

Prefer not to dig through settings? Use a Temporary Chat instead. Open a new chat and click the "Temporary" button in the top corner. That's a chat mode where the conversation keeps no history and isn't used to train the model. One honest caveat: OpenAI says it may keep a copy for up to 30 days for safety review, so treat anything you send as not fully private. Source: OpenAI Temporary Chat FAQ, 2026.

Other chatbots have their own versions of these settings. The list above matters more than any of them, because it protects you no matter which tool you're in or how it's set up.

Step 1: get the text out, minus you

Now the letter. The goal is to hand the AI the confusing part without handing it anything that identifies you. The safest way is to retype just the paragraphs that confuse you. If one of them names a diagnosis or a medication you'd rather keep private, put it in your own general words. Retyping is slower, but you decide exactly what leaves the paper.

If typing is hard on your hands, photograph the letter instead, with one rule: cover your name and every number tied to you, anywhere it appears on the page. Lay a plain sheet of paper over the block at the top that shows your name and Medicare number, then check the footer and any claim lines too, because some notices repeat your Medicare number on every page. A photo carries everything you didn't cover, so when in doubt, the retyped version is the more private path.

What's left after either method is generic benefit language. And generic language is exactly what these tools are good at explaining.

Step 2: the prompt to copy

Here's a prompt you can use word for word:

That last line does the heavy lifting. These tools like to fill gaps with confident guesses, and telling one not to guess cuts most of that off. Paste your retyped text underneath, or attach the fully covered photo, and send it.

Step 3: what comes back, and what it isn't

What comes back is usually genuinely useful: a plain-language translation of the bureaucratic paragraph, a short to-do list with the deadlines pulled out where you can see them, and five questions ready to print and set next to the phone. The question list is the real prize, because it turns a confusing letter into an agenda for a conversation with a real human.

Notice what didn't happen. The tool didn't decide anything. It didn't tell you to keep, switch, or sign up for anything, because you didn't ask it to. That's the frame to hold on to: the AI organizes, it never decides. It's a translator that helps you prepare for a phone call, not an authority on your coverage, and no chatbot should ever confirm a coverage decision.

And if the letter confused you partly because Medicare itself is confusing, our plain-language walkthrough of Medicare Parts A, B, C, and D is the foundation the rest of this sits on.

Where AI gets things wrong, and where you check

These tools can be wrong while sounding completely certain. A 2025 review looked at 89 studies of medical AI answers, and 87.6% of those studies reported incorrect output. Source: Communications Medicine systematic review, 2025.

Medicare has a trap built just for chatbots: the numbers change every year, and these tools learn from older material. Ask one about the drug cap and you may hear $2,000. That was the 2025 figure. For 2026, the out-of-pocket cap in Part D (the drug coverage part of Medicare) is $2,100. Source: CMS, Final CY2026 Part D Redesign Program Instructions, 2026. Same tool, fair question, stale answer, delivered with full confidence. Our 2026 Medicare costs page keeps the current year's numbers in one place.

So the standing rule is simple: the AI explains, the official source confirms. Every number and every deadline it hands you gets checked before you act on it. The checking is free. Medicare.gov has the current numbers and the official version of every rule. 1-800-MEDICARE (1-800-633-4227) puts a real person on the line. And every state has free one-on-one counseling through SHIP at shiphelp.org or 1-877-839-2675. The counselors don't sell anything. Source: Medicare.gov and shiphelp.org, 2026.

One boundary to hold on to. If the question is "what does this letter say," the AI is a fine first stop. If the question is "what should I do about my coverage," that goes to the humans and the official tools, every time.

Questions people ask

Is it safe to put medical information into ChatGPT?

Not personal medical information. General chatbots aren't covered by the health privacy law that applies to hospitals and clinics, and on a standard free account your chats can be used for training unless you change a setting. Source: OpenAI Data Controls FAQ, 2026. Generic language is a different story: once your name, numbers, and dates are stripped out, a confusing paragraph about coverage rules doesn't identify you, and that's the only thing this routine ever shares.

Can AI explain my Annual Notice of Change letter?

Yes, and it's one of the best uses for this routine. The Annual Notice of Change, or ANOC, is the letter your Medicare Advantage plan (Part C, the private all-in-one version of Medicare) or your drug plan must get to you by September 30 each year, listing what changes in your plan the following year. Source: Medicare.gov, Plan Annual Notice of Change, 2026. It lands right before the fall enrollment window, the Annual Enrollment Period, which runs October 15 to December 7. Source: Medicare.gov, Medicare open enrollment, 2026. Redact it the same way, use the same prompt, then take your question list into our ANOC walkthrough and the fall enrollment game plan.

Can AI tell me which Medicare plan to pick?

No, and don't let it try. A chatbot learns from older material, and the plans actually available to you depend on your zip code and change every year. No chatbot should confirm a coverage decision. Plan comparisons happen at Medicare.gov, on the phone with 1-800-MEDICARE, or across the table from a free SHIP counselor. Use the AI to prepare the questions you'll bring to those conversations, and stop there.

Does this work with chatbots besides ChatGPT?

Yes. The routine is the same in any major chatbot: redact first, paste the prompt, check what comes back. The specific settings named on this page are ChatGPT's, so if you use a different tool, spend a minute in its privacy settings looking at how it handles chat history and training. The never-type list doesn't change from tool to tool.

A letter you can finally understand is the win here, but it's the phone call afterward that settles things. When something in your Medicare mail asks you to make a real decision, take your AI-drafted question list to the free official help below, and let a human with the current rules give you the answer that counts.