Proof of Reserves Explained: What an Attestation Can Prove, and What It Cannot

Proof of reserves explained without the marketing spin. This article walks through what a crypto exchange's reserve page, Merkle proof, and accountant-signed report can each actually prove, and what remains unresolved after every one of them passes.

Proof of reserves explained honestly is not a single guarantee. It is a stack of narrow, technical claims, and each one settles a different question. A user can verify their balance is included in a committed dataset. A cryptographic tree can prove that dataset sums correctly. An accountant can attest that specified wallets held specified assets at a specified time. None of those results, individually or together, is the same as an opinion on whether the exchange is solvent or safe.

If you have ever looked at a reserve ratio above 100%, a green "verified" badge, and an accountant's logo, and still felt uneasy about whether your funds are actually safe, that instinct is well-placed. The gap you sense is the gap between what the report technically demonstrates and what the exchange's marketing implies. This article gives you a framework for reading that gap.

Why reserve pages get read as safety evidence

Exchange reserve pages are designed to reassure. They combine a large percentage number, an independent-sounding accountant name, and a self-verification tool that returns a green check when a user pastes their record ID. Those design choices are not accidents. They compress a complex assurance question into something that feels solved.

The confusion is fair. Most readers arrive without a working vocabulary for the underlying professional standards. In everyday English, "attestation" sounds like proof, "reserves" sounds like coverage, and "verified" sounds like verified. In accounting and cryptography, each of those words has a narrow, conditional meaning. This mismatch is where the reader gets hurt. A cautious investor who assumes an "attestation" is an audit, or that a 105% ratio means the exchange has 5% more customer funds than it owes, has been quietly moved past a series of open questions the report itself does not answer. Blockready's approach to platform and counterparty risk sits inside the wider crypto safety framework, and proof of reserves is one specific piece of that map.

The rest of this article decomposes the claim into layers, then hands you a repeatable test you can apply to any exchange's next reserve disclosure. The point is not to defend or dismiss proof of reserves as a category. It is to help you read one clearly, one report at a time.

The Four-Layer PoR Evidence Chain

The clearest way to read a proof-of-reserves system is as four separate proofs, plus a boundary that reminds you what still sits outside the chain. Each layer answers a different question. Each layer can fail, pass, or remain silent independently of the others. And crucially, all four can pass while the exchange remains exposed to failures the layers were never designed to detect.

To make this concrete, picture an exchange that publishes a March snapshot. A user checks their $1,000 balance and the Merkle proof recomputes to the published root. That is Layer 1. The zero-knowledge circuit shows the whole committed dataset sums to $100 million and every balance is non-negative. That is Layer 2. The exchange signs messages from wallets holding $105 million in bitcoin, ether, and stablecoins. That is Layer 3. Reported coverage is $105 million against $100 million, or 105%. That is Layer 4.

What the reader now knows is genuinely useful. This reader's balance was in this exchange's committed dataset. The numbers in that dataset are internally consistent. The exchange demonstrated control of wallets holding at least the claimed asset totals at the snapshot. Reported coverage exceeded reported obligations. Those four statements are more than most non-crypto financial intermediaries publish about themselves in real time.

What the reader still does not know is where most of the risk actually lives. Whether every other customer was included. Whether those wallets were pledged or borrowed elsewhere. Whether the entity that signed the messages was the same legal entity that owes the customers. Whether legal segregation would hold up in insolvency. Whether the assets stayed there after the snapshot. Whether internal controls prevent misuse of customer funds. Whether related-party transactions or affiliate loans quietly recycled capital across balance sheets. Those are separate questions, and none of them lives inside the four proof layers.

What a Merkle tree proves, and where completeness sits

Almost every ranking explainer on proof of reserves reaches the Merkle tree, and then quietly collapses two very different claims. A Merkle proof shows that a specific leaf, meaning your account balance, is included in the dataset the exchange committed to. Altering that dataset would change the published root. That is a clean mathematical result. It is genuinely useful, and it is the foundation of user verification.

The problem is population completeness. A cryptographic proof can guarantee that calculations over the supplied dataset are valid. It cannot, by mathematics alone, prove that the supplied dataset contains every real-world customer, account, product, or legal entity that should have been included. The original academic treatment of proof-of-solvency systems, Provisions by Dagher and colleagues, is explicit about this: cryptographic constructions can prove properties of committed inputs, but management still chooses the inputs. Merkle-sum trees, non-negative balance constraints, and zero-knowledge proofs strengthen the guarantees over the dataset that was submitted. They do not certify that the dataset represents the whole customer book.

User verification does provide one useful safeguard. The more customers who check their leaf, the harder it is for an exchange to systematically omit accounts without eventually being noticed. But this is a statistical detection mechanism at best, not a completeness proof. It works only if enough users check, only after the fact, and only against certain classes of omission. An exchange that excluded an entire product line, an affiliated legal entity, or a category of institutional accounts could still see high user-verification rates from the retail population that was included, because the missing users were never invited to check in the first place.

Asset control is not ownership

Layer 3 is the layer competitors are most likely to overstate. Signed messages, self-sends, and instructed transfers can demonstrate that an entity controls the private keys associated with specified on-chain addresses at a moment in time. Public balances at those addresses can then be checked on a block explorer.

That narrow evidence does not, by itself, establish exclusive possession of the keys, beneficial or legal ownership of the assets, absence of liens or rehypothecation, that the assets were not borrowed just before the snapshot, that the same collateral was not backing another obligation, or that the coins were still there an hour later. Fiat and off-chain assets are harder still: they require bank confirmations, custodian confirmations, and accounting procedures that no blockchain can replace.

This is one of the specific gaps exchange failures have repeatedly exposed. Custody, on paper, is not the same as unencumbered availability, and control at 12:00 UTC is not availability at 14:00 UTC. In a well-run engagement, a practitioner may perform additional procedures to test for encumbrances, obtain confirmations from third parties, or look for evidence of borrowing. Whether any of that happened in a specific report is a question you can only answer by reading the signed report, not the exchange's landing page.

The snapshot problem

Every point-in-time report inherits a timing problem. This is the core criticism from the PCAOB Investor Advisory issued in March 2023, from the SEC Chief Accountant's July 2023 statement, and from most of the accounting profession's neutral commentary since.

A report can be arithmetically correct at its snapshot and still fail to establish ongoing coverage. Reserve assets can be borrowed immediately before the snapshot and returned after. The same collateral can, at different times, back multiple obligations. Customer balances can be moved between products in scope and out of scope. Assets can be sold, lent, pledged, or lost after the snapshot. Even honest, well-run exchanges cannot make the snapshot represent the future.

Higher cadence and synchronized on-chain monitoring reduce some of the risk. A daily or continuous snapshot narrows the window in which a borrowed-asset scheme could stay hidden. Real-time explorer-based checks can catch outflows immediately after the snapshot. These are real improvements over a quarterly report that gives an exchange three months to prepare its numbers. They do not, however, remove the legal, off-chain, and control-side risks that never entered the on-chain view in the first place. Nor do they help against a coordinated scheme in which multiple parties agree to shuffle assets across each other's balance sheets to satisfy overlapping snapshots at different times.

None of this is a hypothetical accusation against any named exchange. It is the structural reason regulators consistently describe reserve reports as informational rather than as substitutes for financial-statement audits.

Attestation, audit, examination, review, and AUP

The single largest source of confusion in this SERP is the word "attestation." In everyday use, it sounds like a guarantee. In professional practice, it is a category that includes several very different engagement types, each with its own assurance level and each governed by its own standard.

Under the AICPA's current SSAE framework in the United States, examinations, reviews, and agreed-upon procedures all fall inside the attestation family. Internationally, ISAE 3000 governs assurance engagements other than audits and reviews of historical financial information, while ISRS 4400 governs agreed-upon procedures as a related service rather than an assurance engagement. And a financial-statement audit, defined for public-company work under PCAOB AS 1000, targets reasonable assurance, high but not absolute, on the financial statements taken as a whole.

The practical implication is simple. If you cannot read, in the signed report itself, which standard was used, what assurance level was expressed, what the exact conclusion or findings were, and what was excluded, then you do not yet know what the report proves. The exchange's marketing page is not the report, and the accountant's name is not the assurance level.

What FTX actually teaches

Almost every mainstream explainer opens with FTX and treats it as a simple proof-of-liabilities story. That framing is convenient but incomplete. According to the Department of Justice's March 28, 2024 sentencing announcement, Sam Bankman-Fried was sentenced to 25 years for orchestrating multiple fraudulent schemes: customer-fund misappropriation, fraud against investors and lenders, and concealment through false records and misleading statements. The failure had many layers, not one.

Any of those layers could have coexisted with a reserve report that looked healthy at some snapshot. A misused customer deposit that had been briefly restored by borrowed assets could pass an inclusion check and a coverage ratio at 08:00 UTC. Related-party loans, off-the-books guarantees, and undisclosed exposures do not appear as line items on a Merkle root. Governance and control failures do not appear on any snapshot. False business records, deliberate concealment from lenders, and improper commingling with an affiliated trading firm involve the kinds of questions a well-scoped financial-statement audit and controls examination are built to ask, and that a narrow point-in-time reserve proof is not.

The correct lesson from FTX is not that proof of reserves would have prevented it, and it is not that proof of reserves is useless. The correct lesson is that customer-fund safety was never reducible to one number in the first place, and that any framework that promises to compress it into one is misleading. Reading a reserve report is one input into an evidence-first evaluation, alongside custody model, legal segregation, governance, controls, and counterparty behavior. This is the same discipline the wider curriculum applies to exchange evaluation and audit-scope literacy: treat every claim as evidence that needs to be sized to its actual scope, and refuse to promote a narrow finding into a broad safety verdict.

A ten-question PoR Report Scope Test

The single most useful thing you can do with any new exchange reserve disclosure is refuse to read the marketing page in isolation. Open the signed report itself, and answer the following ten questions before you form an opinion.

None of these questions is exotic. They are the questions the report itself is designed to answer. If the answers are not obvious or not disclosed, that itself is information. In practice, an exchange whose reserve page cannot survive these ten questions is not automatically dishonest, but its disclosure is thinner than its ratio suggests.

An honest editorial view

Everything above should leave the same practical conclusion. Proof of reserves is neither a scam nor a solvency opinion. It is a family of narrow, useful, but conditional evidence. Better proofs are worth having, and exchanges that publish them are usually more informative than exchanges that do not. But the value of that evidence lives in what the signed report says, not in the size of the ratio or the presence of a badge. A capable reader is safer than a comforted one, and the aim of a good disclosure is to make the underlying question more visible, not more decorative.

The Four-Layer PoR Evidence Chain and the ten-question PoR Report Scope Test presented above are educational reading frameworks, not industry standards. They do not replace the professional standards named in each engagement. Their purpose is to help you notice, on your own, when a marketing page has quietly promoted a narrow finding into a broad safety claim. Proof of reserves is worth taking seriously, precisely because it is worth reading properly.

Frequently Asked Questions