Build Your Own VA Claim System

Veteran Claim System Run your own VA claim Free

JavaScript is off in your browser, so the copy buttons won't work. The download links still do. Use those and open the file in any text editor.

Before we continue, submit an Intent to File

It takes 10 minutes on VA.gov and it's free. It locks your effective date for 12 months, so every week you spend gathering records and getting diagnosed still counts toward your back pay. If you skip it, you're paid from the day you finally submit your claim, which is how veterans lose months of money without ever knowing it.

Do it now, then come back. Submit an Intent to File ›

Nobody can promise you
a rating

I've paid a little over $7,000.
Here's what was actually worth it

$4,882.50 Legal consultant Five separate invoices. It never felt like five thousand dollars until I added it up.

$2,637 Medical opinions Nexus letters and DBQs. Real medical work from real doctors. Worth it.

$0 What finally worked Reading my own records. Knowing what each rating actually requires.

The expensive part wasn't the part that worked. Nobody ever taught me how any of it functioned, so when my claim stalled, I was right back to having no idea why.

Skip ahead and get the zip file

Why this exists

Video coming soon

I served in the Marine Corps from 2008 to 2013. Years later I decided to look into the VA claim system, and I had no idea what I was doing. So I did what most veterans do. I hired a consultant, paid a lot of money, and handed the whole thing over and hoped. It got me some results. Not what I was looking for. And I never learned a thing about how the process actually works.

What I eventually understood was that the missing piece wasn't more money or a better advocate. It was that no record existed of what my life actually looked like day to day. Thirteen years of avoiding doctors, and a long list of problems I'd quietly accepted as just getting older. Nobody had ever asked me about any of it, because nobody had the time.

So I built a system that reads every page of my records, asks me about everything, and shows me the actual regulation behind every rating. It found the path. It also flagged health problems I'd been ignoring that had nothing to do with a rating at all.

Why claims stall

The evidence that decides your rating is the part nobody wrote down

Your rating depends on things nobody ever writes down.

How often it happens
How bad it gets
What you've stopped being able to do

None of that is in your file and it's usually not in your doctor's notes. After 15 years it may not even be in your own memory, because you stopped noticing. You said it was normal because you're getting older.

None of that comes out in a one hour appointment. It comes out when you finally get asked about all of it, question by question, by something that has already read every page of your record.

Setting up

This runs on an AI assistant. Here's how to set one up

The whole system lives inside Claude, ChatGPT, or Gemini. If you've never used one of those, this section is all you need. It takes about ten minutes and you do it once.

If you have never used an AI before

Five things worth knowing first. None of it is technical, and all of it stays true no matter which one you pick.

It reads your documents. It is not searching the internet for your answer. When you point it at your folder it opens your files and reads them. That is where its answers come from, which is why the records matter more than the questions you ask.
It can be confidently wrong. It sounds exactly as certain when it is guessing as when it is right, so certainty tells you nothing. Your tell is simple. This system instructs it to quote the regulation word for word, every time. If it tells you what a rating requires and does not show you the text, do not act on it. Ask to see the text.
You cannot break it and you cannot offend it. Ask the same question five times. Tell it you did not understand any of that. Tell it to start that part over. None of it costs you anything and none of it makes it think less of you.
It only remembers if you go back to the same place. Your folder is the memory. Open a new chat somewhere else and you are talking to a stranger who knows nothing about you or your claim.
Ask plainly. It does what you literally said. "Be thorough" gets you nothing. "Read every page of that file before you answer" gets you something.

Skip ahead and get the zip file

1Open It on a Computer, Not Your Phone

claude.ai, chatgpt.com, or gemini.google.com. Any of the three works. Not the phone app, which can't hold a project together across months.

2Create a "Project" Inside It

Look for Projects in the sidebar and make one called "My VA Claim." Without one, every conversation starts from nothing. Gemini calls its version a Gem.

3Upload the File and Send One Message

Upload the file below, then type: "Follow the instructions in this file." That is the entire setup. It asks about your service a few questions at a time.

4Come Back to that Same Project Every Time

Open the Project, never a fresh chat. It tells you where you left off. Walk away for a month and pick up where you stopped.

The most private option, and the one I use

Keep Your Records on Your Own Machine

Everything above puts your documents into a company's storage, where they sit until you delete them. There is a better way if privacy matters to you. Some versions of these tools read files straight out of a folder on your own computer instead.

Be clear about what that does and does not do. The assistant still runs on the company's computers. When it reads one of your files, the contents get sent to them to be read, every time. What changes is that nothing is stored there: no permanent copy of your records in a project you forgot about, no whole-folder upload, nothing piling up in a chat history. You decide which file gets read and when, and the original never moves.

That also makes the redaction real. Instead of asking you to black things out before uploading, it can strip your identifiers into clean working copies and work from those, while the originals stay put.

Claude. There is a desktop app now, so there are no commands to type. Needs a paid Claude plan, about $20 a month. This is what I use.
ChatGPT. Free. The setup uses a command window once, which the guide walks you through step by step.

Step by step setup guides

Pick whichever assistant you want to use. Each guide covers what that one costs and what its setup involves.

Claude PDF for Mac ChatGPT PDF for Mac

The system works fine without either of these, and most people will start in a browser. But if you would rather your records stayed in a folder you control, this is how.

The flow

Eleven Stages

Your assistant runs these in order. Every single reply it gives you ends with one next step, never a list and never a menu. You don't have to know VA terminology or what comes after this.

1 Protect Your DateIntent to File, ten minutes
2 Privacy SetupBefore anything is read
3 Build The SystemProject and regulations
4 EligibilityEra and exposures
5 Gather RecordsSlowest requests first
6 Read EverythingEvery page, not summaries

See all eleven stages

7 The Questionnaire173 questions, plus yours
8 The MapEvery claim, every gap
9 Get DiagnosedThe words for your doctor
10 Medical EvidenceNexus letters and DBQs
11 File And TrackYou file it, knowing why

Get the File

You'll get a folder called VA Claim with the toolkit already inside it and a START HERE page that walks you through the rest, step by step.

Download the zip

Prefer to not deal with a zip? The plain text file is below, along with the same content split into its two halves.

Part 1: Orientation An orientation guide written for you. It includes the process, red flags, glossary, and privacy information. Download
Part 2: Setup prompt No need to read it if you don't want to. These are the instructions written for your AI. Download

What you get

Not a Chatbot That Guesses

The Regulations, Quoted

It downloads the actual federal rating schedule and quotes it, rather than somebody's summary of what a rating requires.

A Gap Analysis Per Condition

The criteria for the next bracket, what you currently meet, and the one piece of evidence that would close it.

A Questionnaire From Your Chart

Built after reading your records, so it asks about what you've normalized and about findings nobody followed up on.

Health Flags, First

The same reading finds things that need a doctor for your health, not your rating. It's told to raise those ahead of any claim analysis.

The details

The parts you'll want before you start

It Said Something That Sounds Wrong

Trust that feeling. You know your own life and your own body better than it does. It is reading documents about you; you were there. When those two disagree, you are usually the one who is right.

If it is about a rating or a rule: ask to see the text. "Show me the actual text of that regulation." It is instructed to quote the regulation word for word rather than summarise it, so if it cannot produce the text, it did not read it. Do not act on it.

If it is about your history: tell it plainly that it has something wrong and what the truth is. "That is not right. It was 2011, not 2014, and it was my left knee." It will correct the record. Do not let a wrong detail stand because you did not want to argue with it.

If a document it wrote says something you do not recognise about your own life, that sentence does not go to VA. Not a softened version of it, not "close enough". One invented or overstated detail can damage how every other condition in your file is read.

If it keeps getting the same thing wrong, ask it to check its own work: "Check what you just told me against the documents in this folder and tell me what you were unsure about."

Something Isn't Working

Your assistant won't read the whole file. Tell it: "Read the entire file from the beginning to the end before you reply." Some tools skim long documents by default. If it still won't, split the file and give it the Orientation and the Setup Prompt separately.

It's summarising the regulation instead of quoting it. Push back: "Show me the actual text of that regulation." The system is built to quote rather than paraphrase, and if it cannot show you the text, it has not read it. Do not act on a rating requirement you have not seen quoted.

It forgot everything since last time. You are probably in a new chat rather than in the Project. Open the Project itself, then start a message there.

It stopped partway through your records. Free tiers cap how many documents you can upload in a day. Wait for the reset, or feed it fewer files at a time. It works, it is just slower.

It asked you for the same thing twice. Tell it to check the files already in the Project before asking. Long projects drift.

Still stuck, or something on this site is broken? Email help@vetclaimsystem.org. Do not send medical records, claim numbers, or personal information. I cannot review your file, and I am not your representative. I read everything and answer when I can.

How the Rating Numbers Actually Work

Combined ratings aren't addition. VA combines them, so they never simply add up. At 90%, a new 30% rating moves you about three points, not thirty. Knowing that before you chase a condition saves you months spent on the wrong one.

Ratings also lock over time. Three clocks matter, and your assistant tracks all three:

Five years. A rating held this long is treated as stabilized, and VA has to clear a higher bar to reduce it.
Ten years. Service connection itself can no longer be severed, short of proven fraud.
Twenty years. The rating percentage is protected for life and cannot be reduced below that level.

These are the reason a claim you filed years ago can be worth more than a new one, and most veterans are never told they exist.

More on Setting Up

Why not the phone app. Phone versions can chat, but they can't hold a project together across months or manage a folder of your records. This runs for a long time, so start on a laptop or desktop. The desktop apps are fine too.

What a "Project" actually is. There's no special mode, no separate app, nothing to install. It's the same website you'd use to ask any question. A Project is just a container holding your files, your history, and the instructions it works from. Menus move around, so if you can't find it, ask the assistant "where do I create a Project?" and it will walk you through it.

You don't have to read the file. Don't open it, edit it, or fill anything in. Upload it, send the one message, and it takes over from there. It builds its own filing system and does the organizing itself.

Which AI Do I Need?

Whichever you already have. Claude, ChatGPT, and Gemini all clear the bar. It's a text file, not an app, and nobody here is paid to recommend one. What matters is that your tool can:

Read a lot of documents completely. Records run to hundreds of pages. This is the requirement that separates them most.
Remember your project between sessions. This runs for months, so look for a "Projects" feature.
Fetch a web address. The government source for the rating schedule is already written into the prompt.
Run code, so combined ratings are computed rather than estimated.
Give you files back, like the questionnaire and your tracking documents.

⚠️ Free tiers are where you'll hit a wall. Some cap you at a few document uploads per day. Workable, just slower.

What Happens to My Medical Records?

Nobody here ever sees them. There's no account, no upload to this site, no server on our end. You're working in your own AI account, with documents that already belong to you. Whatever happens to your records is between you and the tool you choose, which is why it's worth choosing deliberately.

The most protective setup is an assistant that reads files off your own computer. Your records stay on your machine and are never uploaded anywhere. That's a property of how those tools work, not a promise anyone has to keep.

If you upload to a cloud service instead, that service has received your documents. Nothing can un-send them. That's the honest tradeoff for the convenience, and it's worth knowing before your first upload rather than after.

The setup prompt tells your assistant to strip identifiers from working copies and to keep your name, Social Security number, file number, and address out of anything it generates. Understand what that is: an instruction, not a guarantee. These systems usually follow instructions and sometimes don't. So the real protection is you. Read anything before you submit or share it, and check your identifiers aren't sitting in it.

Where Do the Rules and Ratings Come From?

Nothing here is anyone's interpretation. Every criterion your assistant quotes is downloaded verbatim from the government's own published regulations, and you can read the same sources yourself.

eCFR, Title 38	The complete rating schedule. Part 4 is what conditions are rated under. Part 3 covers service connection and how ratings become permanent.
govinfo, GPO	The Government Publishing Office's official copy, as PDFs. First fallback.
Cornell Law	A long-running mirror from Cornell Law School. Often easiest to read on a phone.
Board decisions	Every past Board of Veterans' Appeals decision, searchable. Useful for seeing how criteria get applied.

Your assistant tries all three regulation sources in order and verifies that what it downloaded is actually the regulation, because a page can return "success" and contain nothing. If every route fails it's told to stop and ask you to upload the page rather than quote something it couldn't read. If it ever tells you what a rating requires without showing you the text, push back.

What This Is Not

Not legal advice, and not a representative. An AI can't represent you, and no one can promise a rating or a faster claim.
Not a way to get a rating you don't deserve. It's built the opposite way, to catch invented findings and overstated symptoms before they reach VA. One fabricated detail can cost you credibility on every legitimate condition in your file.
Not something that files for you. Everything gets your own eyes first. If a generated statement says something you don't recognize as true about your life, it doesn't go in.
Not finished in a weekend. Testing gets scheduled, records take months, diagnoses take appointments. This is a system for a process that runs in years.

Veteran Claim System, build your own VA claim system