cryptobank
Crypto Bank is an independent editorial publication. It is not a bank, is not a deposit-taking or custodial institution, holds no banking, trust, money-transmitter or e-money licence, and offers no FDIC or FSCS protection. Nothing here is financial, legal or investment advice.
BTC $63,224-0.77% ETH $1,773-0.88% USDT $1.00+0.10% USDC $1.00+0.01% SOL $73.99+2.41% XRP $0.98+2.53% BNB $584-0.40% ADA $0.15+0.59%
Proof of Reserves and What It Cannot Prove

GUIDE

Proof of Reserves and What It Cannot Prove

Crypto Bank Editorial Team · November 17, 2026

“Proof of reserves” sounds like it settles the question of whether a provider has your coins. It does not, and the gap between what the phrase suggests and what the underlying technique can actually show is wide enough that it is worth walking through mechanically, rather than taking the label on faith.

What a proof of reserves actually checks

The common technique is a Merkle-tree inclusion proof. The provider takes a snapshot of every client balance, arranges them into a Merkle tree (a structure that lets each individual balance be hashed together with others up to a single root hash), and publishes the root. An individual client can then check that their own balance was included in the tree that produced that published root, without the provider having to reveal every other client’s balance publicly. Separately, the provider (or an outside reviewer) adds up the total client balances claimed and compares that total against the coins actually held in wallets the provider controls, verified by on-chain signatures.

Done correctly, this shows two real things: that a specific client’s claimed balance was part of the set used to compute the published total, and that the provider controls on-chain wallets holding at least that published total in assets. Both of those are genuine, checkable facts. The problem is what sits outside that scope.

What it cannot show: liabilities

A proof of reserves demonstrates assets. It says nothing about liabilities — what the provider owes, to whom, and under what terms. A provider could hold a Merkle-tree-verified pool of coins equal to 100 percent of client claims, while simultaneously owing a separate, undisclosed debt to a lender, an insider, or another part of its own corporate structure that would make it insolvent the moment that debt is called. The Public Company Accounting Oversight Board has noted that these snapshots are inherently limited and should not be read as evidence that an exchange’s assets are sufficient to meet its liabilities — the reserve side of the ledger is simply not what the proof is measuring.

This is the single most cited criticism of proof-of-reserves exercises, and it is also the most mechanically straightforward one: a balance sheet needs two sides to mean anything, and a standard proof of reserves publishes one.

What else a snapshot misses

  • Encumbrances. The on-chain coins a provider controls at the moment of the snapshot may be pledged as collateral elsewhere, borrowed against, or subject to a lien the proof does not and cannot reveal.
  • Point-in-time only. A snapshot shows holdings at one moment. It says nothing about balances the day before or the day after — a provider could temporarily borrow assets to pass a snapshot and return them afterward, something more than one critic has flagged as a real manipulation risk rather than a theoretical one.
  • Off-chain and non-crypto assets. Fiat balances, receivables, and other off-chain claims generally are not part of a Merkle-tree proof at all, even though they may be a meaningful share of what a provider reports owning.
  • Exclusivity of control. Signing a message from an address proves the signer currently controls that address’s keys. It does not prove no one else also has a copy of those keys, or that the keys were not borrowed for the exercise.
  • Methodology and scope choices made by the provider itself. Which wallets are included, which client accounts are included, and how “liabilities” are defined (if defined at all) are choices the provider makes when designing its own proof — there is no single, standardized template every provider follows. This is why two providers’ proofs of reserves are rarely directly comparable, even when both call the resulting document by the same name: one may include every client balance and every supported asset, while another covers only its largest coins or excludes accounts flagged as dormant, without either choice being stated prominently.

How this relates to attestations and audits

A Merkle-tree proof of reserves is sometimes paired with a third party’s attestation report confirming that the provider’s claimed total matches the wallets checked at that moment. That pairing can add some assurance that the numbers were not simply made up, but an attestation of this kind is still a point-in-time procedure with a scope the provider and the reviewer agree on in advance — it is not the same thing as a financial statement audit, which tests controls and liabilities over a period of time rather than a single moment. The distinction matters enough that we cover it in its own guide: attestation vs. audit, the difference that matters.

Why providers publish them anyway

None of this makes a proof of reserves worthless, which is worth stating plainly given how much of this guide is about its limits. In the period following the collapse of several major exchanges that turned out to be insolvent while still operating normally, “we will never publish a proof of reserves” became a far more alarming signal than publishing an imperfect one. A Merkle-tree proof is a genuine, verifiable fact about asset control at a point in time, and a provider willing to let individual clients check their own inclusion is disclosing more than a provider that offers no mechanism for clients to check anything at all. The honest framing is that a proof of reserves is a floor, not a ceiling: it rules out the specific failure mode of a provider having no matching assets whatsoever, while leaving open every other way a custodian can still fail.

The most widely discussed real example remains the exchange proofs published with the assistance of outside accounting firms in the wake of the FTX collapse. Commentators at the time, including analysts at major banks, pointed out that even a well-executed version of this exercise proved reserves and nothing else — the criticism was not that the Merkle-tree math was wrong, but that market participants and headlines treated “proof of reserves published” as equivalent to “fully solvent and safe,” which the underlying exercise never actually claimed.

How to read a proof-of-reserves disclosure, practically

  1. Check whether the publication names a reviewer and describes the review’s scope, or whether it is purely the provider’s own, unreviewed snapshot.
  2. Look for any statement about liabilities at all. Its absence is itself informative, not neutral — it means you are seeing one side of the ledger only.
  3. Check the date and frequency. A single proof published once, long ago, tells you less than a reserve-and-liability disclosure repeated on a regular cadence.
  4. Check whether client asset types, not just a single headline coin, are covered. A proof covering bitcoin reserves says nothing about whether other asset balances are backed.

We track what providers have published on this front, with the date and the reviewer named where one exists, on our reserves and attestation tracker.

Where this fits the Custody Question

A proof of reserves is mostly relevant to the first row of the Custody Question we run on every provider profile: who holds the keys. A credible proof is evidence that the provider does, in fact, control the on-chain assets it claims to. It has little to say about the second row (who insures it) and almost nothing to say about the third (what happens on default) — that third question is a legal one, decided by contract terms and insolvency law, not by a blockchain signature.

FAQ

Is a proof of reserves the same as an audit?
No. A standard proof of reserves is a point-in-time, assets-only exercise, often performed without a named independent reviewer. An audit tests controls and both sides of the balance sheet over a period, performed by an independent auditor under a recognized standard.
Can a provider fake a proof of reserves?
A provider can borrow assets temporarily to pass a snapshot, choose a narrow scope that excludes problem wallets, or simply not disclose liabilities — none of which requires falsifying the cryptography itself. The inclusion-proof math can be entirely sound while the overall picture it supports is still misleading.
If a provider has never published a proof of reserves, does that mean something is wrong?
Not necessarily, but it does mean there is less public evidence either way. “No disclosure found” is a fact worth noting on its own, not proof of a problem, and not proof of safety either.
Does a proof of reserves prove the provider is not using client assets for its own trading?
No. A snapshot showing adequate reserves at one moment says nothing about what the provider does with those assets between snapshots — lending them, using them as trading collateral, or otherwise putting them at risk is a separate question the proof does not address, and is closer to the segregation question covered in our segregated vs. omnibus accounts guide.

Risk note

Any specific proof-of-reserves exercise described by a provider should be checked against its own published methodology and date, not assumed from general descriptions like this one. A past proof of reserves is not a current guarantee.

Not a bank. Not a deposit-taking institution. No FDIC, FSCS or deposit-insurance cover. Editorial only.

Not financial advice.