Proof-of-reserves pages became common after several custodians and exchanges turned out to hold far less than customers believed. The best versions of these pages use real cryptography to let an individual user check their own balance is included in a snapshot; the weakest versions are a number on a page with no way to check anything. Telling the two apart is a skill worth having before trusting either.
The idea behind a merkle-tree reserves proof
A merkle tree is a way to commit to a large set of data (every customer’s balance, in this case) with a single short value called a root, such that anyone holding one item from the set and a short “proof path” can confirm that item was part of the committed set, without needing to see every other customer’s balance. A custodian publishes the root; each user can retrieve their own leaf and proof path (usually through an account dashboard) and run an open-source verification tool to confirm their own balance was actually included in the root that was published. This lets thousands of individual users each confirm their own piece, collectively making it very hard to quietly exclude or understate anyone’s balance without the root changing in a way the custodian would have to explain.
Step by step: checking a proof-of-reserves claim
- Find the published merkle root and the snapshot date and time it corresponds to. A proof without a stated snapshot time is not a check you can run.
- If you hold an account with that custodian, look for a self-serve inclusion check – typically a page where you log in and the system shows your leaf data and proof path, or lets you download them.
- Run the inclusion proof, ideally with an independent, open-source verifier rather than only the custodian’s own tool, to confirm your balance hashes up to the published root.
- Separately, look for an independent on-chain reserves check: many custodians publish the wallet addresses they hold assets in, which lets anyone (not just account holders) read the actual on-chain balances against the total liabilities figure the custodian claims, using the block-explorer reading skills from how to read a block explorer.
- Note who performed any attestation of the process (see how to find and read a stablecoin reserve attestation for what an attestation is and is not) and whether that report is dated at or near the snapshot, not months earlier.
What proof of reserves does and doesn’t prove
A well-built merkle proof, checked by enough individual users, makes a real claim: at a specific snapshot moment, the published root correctly committed to a set of customer balances, and assets were held somewhere matching (at least) the total of those balances. What it does not show:
- Liabilities beyond customer balances. A reserves snapshot typically covers customer-owed assets, not other obligations a custodian may carry (loans, bonds to counterparties, legal liabilities).
- Continuity between snapshots. Reserves can be borrowed just before a snapshot and returned immediately after – sometimes called “window dressing.” A single point-in-time proof says nothing about balances the day before or the day after.
- Asset quality or liquidity. A reserve can technically be present but illiquid, encumbered, or itself a claim on a third party rather than directly held. A snapshot number does not distinguish these without additional disclosure of what the reserve assets actually are.
- Whether the proof tool itself is honest. If the only verifier available is the custodian’s own closed tool, you are trusting their code to tell you the truth about their own books, which narrows the independence the whole mechanism is supposed to provide.
Cross-checking with on-chain wallet disclosures
Some custodians go a step further than a merkle-tree inclusion proof and publish the actual wallet addresses holding customer assets. When that disclosure exists, it opens up an independent check that doesn’t depend on trusting the custodian’s own proof tool at all: using the block-explorer reading skills from how to read a block explorer, anyone can look up the disclosed addresses directly and read their real, current balances from the chain itself, then compare that total to the liabilities figure the custodian claims to owe customers. This is a meaningfully stronger form of transparency than a merkle root alone, because the underlying wallet balances are independently and continuously checkable by anyone, not only by customers running an inclusion proof against a single historical snapshot.
The caveats still apply, though: a custodian can disclose only a subset of its wallets, can move assets between disclosed and undisclosed wallets around the time of a check, and the mere existence of a large on-chain balance in a disclosed address doesn’t tell you whether that balance is encumbered by a loan, pledged as collateral elsewhere, or otherwise not as unconditionally available as it appears. Reading the balance is the easy part; reading what claims might sit against that balance is the part that usually isn’t visible on-chain at all, and has to come from other disclosures.
Not financial advice
This guide explains how proof-of-reserves cryptography works and how to check an individual inclusion proof. It is general education, not financial, legal, or investment advice, and a passed proof-of-reserves check at one point in time does not guarantee an exchange or custodian’s ongoing solvency.
Comparing how different custodians approach this
Not every custodian that publishes something called “proof of reserves” uses the same method, and the label alone doesn’t tell you which version you’re looking at. Some publish a full, individually checkable merkle-proof system; others publish only a periodic accountant’s attestation of a total balance with no per-user inclusion check at all; and some publish little more than a dashboard number with no supporting mechanism or report behind it whatsoever. Treating all three as equivalent because they share the same marketing phrase is a mistake worth avoiding – the actual verification mechanism behind the phrase, not the phrase itself, is what determines how much weight the claim deserves.
Why how often a proof is published matters
A custodian that runs this process once, publicizes it widely, and never repeats it has given you one historical data point and nothing about its current state. A custodian that repeats the process on a regular, stated cadence – and keeps a visible archive of past snapshots rather than only showing the latest one – gives a reader a track record to check against, which is a meaningfully stronger signal of an ongoing commitment to transparency than a single, one-time proof ever can be, no matter how rigorous that single proof was.
FAQ
Can proof of reserves prove a custodian isn’t insolvent?
No. It can show that assets matching customer balances existed at a specific snapshot. It cannot show total liabilities, including non-customer obligations, which is what solvency actually depends on.
What if a custodian only publishes a total number with no merkle proof?
Then there is nothing for an individual to check. A headline reserves ratio with no inclusion proof and no on-chain wallet disclosure is a claim, not a verification mechanism.
Does checking my own inclusion proof tell me anything about other customers?
Directly, no – your proof only confirms your own leaf. Indirectly, because many customers can each check their own leaf against the same published root, a root that excluded or understated balances would require those customers’ checks to fail, which is the collective strength of the mechanism.
Why does the snapshot date matter so much?
Because reserves can move. A proof tied to a specific, recent timestamp tells you about that moment; an old or undated snapshot tells you very little about current holdings.
Not financial advice. Human-edited; AI-assisted drafting is disclosed in our editorial guidelines. Spotted an error? See how corrections work.


