Built for one country's rules
The strongest appeal a U.S. practice can send starts with the most complete evidence base behind one in the country. That is what Merits is — assembled for American denials, American plans, and American law, and nothing else.
Why a closed library, not an AI that “knows” the rules.
A general model will write a confident appeal in seconds, then cite a regulation that does not exist. On a letter going to a payer under a doctor's name, that is the entire risk.
So Merits never asks a model what the rules are. It retrieves them — from the authority that published them, captured word for word, re-checked on every citation — and walls off the writing so a letter can quote only what is actually on the shelf. A source that isn't in the library cannot reach a letter. Inventing one is not frowned upon; the architecture makes it impossible.
What is on the shelf
The library, itemized and verified.
- 484payer medical policiesThe plan's own coverage criteria, across 43 insurers — quoted back to the payer that wrote them.
- 40,000+external-review rulingsPublic independent-review decisions — what actually overturned denials like the one on the desk.
- 240FDA drug labelsIndications and Usage, verbatim — the strongest anchor a medical-necessity appeal has.
- 27FDA device approvalsPMA, 510(k) and De Novo grants for the high-dollar procedures — TAVR, MitraClip, spinal-cord stimulation.
- NEJM+ JAMA landmark trialsThe pivotal studies the payers themselves cite, paired to the procedures they support.
- 4,307codes mapped to Medicare LCDsLocal coverage criteria, fetched live for the exact procedure code on the claim.
- 12federal texts, hash-verifiedERISA, ACA and Medicare appeal law, captured verbatim from eCFR and the U.S. Code.
- 1.7MNCCI code-pair rulesThe bundling edits behind every CO-97 and CO-4 denial, matched by exact code pair.
- 51jurisdictions mappedExternal-review rights and filing deadlines for every state, plus D.C.
The layers stack. A denied MitraClip does not get a form letter — it gets the FDA approval for the device, the New England Journal trial behind it, and the payer's own published criteria for the procedure, quoted back to the payer that wrote them.
The honesty that costs us money.
A tool paid by the letter has every incentive to call every denial winnable. Merits flags the ones that are not — before payment, at no charge — and if it decides mid-draft that no honest appeal exists, it returns the credit. Turning down the money is the only version of this a physician can actually trust.
The charter
What this instrument refuses to do.
Part of what makes an appeal worth trusting is what the tool will not let happen. Six refusals are load-bearing.
- I
It cannot see who the patient is.
Name, member ID, date of birth — stripped in the browser before anything is sent, and restored on the device at download. What is never received cannot leak.
- II
It cannot cite outside the library.
Writing is walled to the closed, source-verified index. A source that isn't on the shelf cannot reach a letter. Inventing one is not discouraged here; the architecture makes it impossible.
- III
It will not deliver a letter it can't support.
Every quoted passage is checked word-for-word against its source. An unsupported clinical claim is pulled and handed back as a flag; a draft that fails validation is refused, not shipped.
- IV
It will not charge for a losing fight.
Denials that rarely win on the merits are flagged before payment, at no charge. If the engine decides mid-draft that no honest appeal exists, the credit returns.
- V
It will not make the clinical call.
The letter stays a draft until it is reviewed, every flag resolved, and signed. It goes out under the provider's name, on the provider's judgment — and never names Merits or AI.
- VI
It will not lock anyone in.
No contract, no per-seat fee, no account to begin. Pay per denial when one is worth the fight, or a flat month when appealing is routine.
HELD TO ON EVERY LETTER — FROM THE FIRST DRAFT TO THE ONE YOU SIGN.
The method
How a letter is built — upload to signature.
- 01
Read.
The denied EOB is read in the browser. Patient identifiers are stripped on the device — the de-identified claim lines are all that is ever sent.
- 02
Classify.
The denial code and plan type route to a strategy. A bundling denial argues NCCI edits; a medical-necessity denial argues coverage criteria; an ERISA plan gets federal law, never state law.
- 03
Assemble.
Evidence is retrieved from the closed library for the exact code and payer, and the argument is drafted around what the sources actually say.
- 04
Verify.
Every quoted passage is checked word-for-word against its source. Unsupported claims become flags; a draft that fails is refused. Then it waits for the provider's judgment and signature.
In the open
Its limits, in writing.
A standard is only worth the name if it says where it ends. So this is the model card — what Merits will not claim, and what it does not do. It is published on purpose.
No win-rate claims
The outcome data shown is public independent-review data, labeled as such. There is no published Merits success rate — a figure like that says more about which denials people choose to appeal than about the tool, and a borrowed one isn't ours to show.
Not every denial
Pharmacy-benefit drug denials — the PBM lane — are out of scope today. Medical-benefit drugs, procedures, imaging, devices, and administrative denials are in.
A draft, not a guarantee
A well-cited appeal materially strengthens the record, and most appealed denials nationally do get paid. No letter can promise an outcome, and this one never pretends to.
Not legal advice
Merits assembles the regulatory and clinical record. It is not a law firm, and the letter is not legal advice — for a litigation-track dispute, bring in counsel.
Who answers for it
Accountable for every rule in it.
Its incentives answer to the provider, not the payer — not an insurer's side project, a billing company's upsell, or a private-equity roll-up. Every rule in it is owned, and every rule is answerable.
A letter a physician signs their name under should be reachable and answerable. Write to [email protected] — questions and corrections reach a person, not a queue.
What touches your data
Stored little. Deleted fast.
- Patient identityNever sent — placeholders are substituted in your browser and restored on your device at download.
- The uploaded EOBDeleted from storage once generation completes.
- Payments & logsCarry a case ID only. Nothing clinical rides through Stripe or email.
- InfrastructurePayments by Stripe; database, storage, hosting and drafting run on vetted providers under strict agreements, including no-training terms for the AI infrastructure.
The full policy, in plain language: meritsappeals.com/privacy
Straight answers
The questions worth asking us.
- Is Merits legitimate?
- Merits turns a denied EOB into a cited appeal letter, built on U.S. clinical and legal sources. Every citation links to its public source — CMS, eCFR, FDA, the payer's own published policy — so the work can be verified before it's signed. A complete sample letter is at meritsappeals.com/samples.
- How does Merits work?
- The denied EOB is uploaded. Patient identifiers are removed in the browser. The denial is classified against a strategy library, evidence is retrieved from a closed source-verified index, and a letter is drafted in about a minute — with every quote checked verbatim against its source before anyone sees it.
- Does Merits store patient data?
- Patient identity never reaches Merits servers: names, member IDs and dates of birth become placeholders in the browser and are restored on the device at download. The uploaded EOB is deleted after generation.
- Does Merits guarantee the appeal will win?
- No — and no one honestly can. What is guaranteed is the record: the right argument for the denial type, tied to sources anyone can open, filed inside the deadline. When a denial is rarely winnable, Merits says so before payment and doesn't charge.
- Who is behind Merits?
- Merits answers to the provider, not the payer — not an insurer's side project, a billing company's upsell, or a private-equity roll-up. Every rule in it is owned and answerable, at [email protected].
- What does Merits cost?
- $9 per letter, pay-as-you-go, with volume credits down to $5 — or $20 a month for 20 letters when appealing is routine. Checking whether a denial is worth appealing is always free.
So the people who treat patients keep what they've earned.
If a denial crossed your desk that you were about to write off, start there.