Anonymous vs pseudonymous: what Bitcoin actually records

Linkability is the property that two identifiers belong to the same party. Bitcoin does not publish that fact, but it publishes enough structure for an observer to infer it. The best-known inference is the common-input-ownership heuristic: when a transaction spends several outputs as inputs, the observer assumes that one party controls all of them, because producing the signatures requires the keys for every input. That assumption is not a protocol rule and it is not always true — a collaborative transaction such as a CoinJoin deliberately spends inputs from several parties — but it is true often enough that it forms the basis of commercial chain analysis.

Once two identifiers are linked, the link is transitive in the observer's model. If address A is linked to address B by one transaction and B is linked to C by another, the observer treats A, B and C as one cluster, and every payment any of them makes is attributed to that cluster. The cluster is not a person, but it is a behavioural unit, and it is the unit that analysis firms sell. A single careless transaction can therefore merge two sets of holdings that the owner had deliberately kept apart, and the merge is permanent.

Linkability is also what makes the distinction between anonymity and pseudonymity concrete rather than philosophical. If a cluster is never connected to a person, the pseudonym has held. If it is connected once — through a withdrawal, a purchase, a published address — then every transaction in the cluster is connected, not just the one that made the link. The tracing page sets out the heuristics in detail, and the limits of analysis page covers where the inference breaks down.

The first common error is to treat "no names on the ledger" as anonymity. It is not, because the ledger is not the only source of information. An observer who can match one address to a person can read the rest of the cluster, and the matching is often trivial: exchanges are regulated, merchants keep records, and addresses are frequently published on purpose. The protocol's silence about identity is a design property, not a guarantee about the world around it.

The second error runs the other way and treats Bitcoin as fully traceable. That overstates the case. The heuristics are probabilistic and they produce false positives; mixing, custodial flows and off-chain settlement all break the chain of inference; and a cluster that is never linked to a person remains a set of pseudonyms. The honest position is that Bitcoin is transparent by default and that privacy is achievable through deliberate practice, not that it is either guaranteed or impossible.

The third error is to assume that a fresh address is a fresh identity. Addresses are free and unlimited, but they are not independent: a wallet that spends two of them together has published the link, and a wallet that derives them in sequence from one seed has published a structure that an observer can sometimes exploit. The address reuse and change addresses pages take each of those cases in turn.