Rating Integrity

What we will not do

Some tools in this market sell a number. They promise to get you to 100 percent, and the only way to deliver that promise for everyone is to encourage people to overstate what is wrong with them. We do not do this, and specifically:

• We will not script your C&P exam. Our guides explain how the exam works and what the examiner is measuring, so you can walk in prepared. They do not tell you what to say, choose which day to describe, or hand you wording. Your answers should describe your actual condition. Do not exaggerate, and do not tough it out and downplay real limits either, because minimizing is just as inaccurate and it costs veterans real ratings.
• We will not add conditions you do not have. When we surface a potential secondary condition, we are pointing at a documented medical relationship for you to discuss with a clinician. It is a question to raise, never a condition to claim because it might pay.
• We will not promise an outcome. VeteranHQ does not prepare, file, or represent VA claims, and nothing here is legal or medical advice. Every rating figure we show is an estimate of what the criteria support. The VA decides your rating.
• We do not frame your claim as an optimization problem. You will not find us ranking conditions by payout or telling you which one gives the best return. That framing pushes people toward claiming for money rather than for what is actually wrong.

Using VeteranHQ to prepare, support, or submit a fraudulent or exaggerated claim is a violation of our Terms of Service and grounds for termination of your account.

The steps we take to calculate your rating

Every rating figure on VeteranHQ is produced by the following process. The design goal throughout is that the number we show you is understated rather than overstated, so that what we tell you is defensible with the evidence you actually have.

1. Step 1

   We start from the regulation, not from a target number

   Ratings are evaluated against the criteria in 38 CFR Part 4, the VA's own rating schedule, at the diagnostic code that governs the condition. We derive the criteria from the regulation text before writing any code for it, and we cite the specific section so you can read it yourself. We never work backward from a percentage someone wants.

2. Step 2

   The rating comes from a rule, not from an AI

   The step that decides a percentage is ordinary deterministic code evaluating the rating schedule. No language model is involved in that decision. This is a deliberate architectural choice: a rule evaluation is reproducible and auditable, and it cannot invent a percentage the way a language model can. Where we do use AI, it reads documents to pull out clinical facts, and it is never the thing that picks your rating.

   The AI that does read your records runs under a signed Business Associate Agreement with Zero Data Retention enabled. Your prompts and the model's outputs are not stored by the vendor, and your records are never used to train anyone's model. PHI is architecturally prevented from reaching any provider we do not hold a BAA with. Our security page lists every subprocessor that touches your data and the date each agreement was executed.

3. Step 3

   Every clinical fact has to be traceable to a quote

   When we read a fact out of your medical evidence, that value must carry the exact verbatim passage it came from, and we verify the quote really appears in the document. If a value cannot be traced to a supporting quote, it is discarded rather than used. A fact that cannot be pointed at in your records is not a fact we will rate you on.

4. Step 4

   When your records disagree, the answer stays unknown

   Records contradict each other all the time. If one note says a symptom is present and another says it is absent, we do not pick the version that produces a higher rating, and we do not let confidence scores break the tie. The field is left unknown. An unresolved contradiction has to understate rather than guess, because guessing in your favor is exactly how a tool starts inflating ratings.

5. Step 5

   We show you the floor and the ceiling, not a single flattering number

   Every estimate is computed twice. Once with the unknowns left out, which gives the rating your evidence proves today. Once with the unknowns resolved in the most favorable direction, which gives the rating that is reachable if you document those things. An unknown can only raise the ceiling, never the floor. If the two numbers differ, we show the range and tell you exactly which missing facts would close the gap, which turns the gap into evidence to go gather rather than a number to assert.

6. Step 6

   Combined ratings use the VA's table, not addition

   Multiple conditions combine under 38 CFR 4.25, which is not addition. We arrange your ratings from most to least severe, combine them pairwise through the regulation's Combined Ratings Table, round to a whole percent after each step as the table requires, and only convert to the nearest 10 at the very end. Where the bilateral factor under 38 CFR 4.26 applies, we apply it.

   Doing this exactly matters in both directions. A shortcut that combines everything in one pass instead of rounding at each step underrates the veteran by a full 10 points on 145 of the 7,997 possible rating combinations. You can check our arithmetic yourself: the combined rating calculator shows every intermediate step rather than just a final answer.

7. Step 7

   It is an estimate, and we label it as one

   Everything above produces an estimate of what the rating criteria support given the evidence in front of us. It is not a decision, not a prediction, and not a promise. The VA rates your claim based on the full record, including its own examination. We mark every estimate as an estimate rather than dressing it up as a result.

Why an inflated rating is not a win

Beyond the criminal and civil exposure that comes with defrauding a federal benefits program, the regulations single out fraud as the thing that strips away protections you would otherwise keep for life.

• A rating held continuously for 20 or more years normally cannot be reduced below that level. Under 38 CFR 3.951(b), it can be, on a showing that the rating was based on fraud.
• Service connection in effect for 10 or more years normally cannot be severed. Under 38 CFR 3.957, it can be, on a showing that the original grant was based on fraud.

The protections that make a rating durable are exactly the protections fraud voids. A veteran who overstates one condition puts the ratings they genuinely earned at risk alongside it. Accuracy is not just the honest path here, it is the one that holds.

What we do instead of inflating numbers

Most veterans we see are underrated, not overrated, and the reason is almost never that they were insufficiently aggressive. It is that a real, service-connected limitation never made it into the record. So the work we do is documentation, not persuasion: find the condition that was never claimed, find the secondary nobody connected, find the evidence gap that is holding a claim at a lower tier, and help you fill it with real medical documentation.

That approach gets veterans to accurate ratings, and accurate ratings survive review.