SWIFT is a messaging network, not a settlement system. That distinction matters more than you think

SWIFT moves messages, not money. What that means when a payment is delayed, what gpi and the ISO 20022 migration changed, and what they did not.

4 mins
August 21, 2026

What SWIFT actually does

"Send it via SWIFT" is the common shorthand for an international payment, and it is a misleading one.

SWIFT does not send payments. It sends messages about payments. The distinction sounds pedantic until a payment is delayed, and the person asking where their money is discovers that the two questions — where is the message and where is the money — have different answers.

This post sets out what SWIFT is and is not, traces the separation between the message and the settlement it describes, and explains what changed and what did not when SWIFT retired its legacy MT format in November 2025.

SWIFT, the Society for Worldwide Interbank Financial Telecommunication, is a member-owned cooperative headquartered in Belgium. It was founded in 1973 to replace telex as the way banks exchanged payment instructions, and it now connects more than 11,500 institutions across over 200 countries, exchanging tens of millions of standardised messages a day.

What it provides is the format and the wire, not the money. A SWIFT message carries the details of a payment instruction: the amount, the currency, the ordering and beneficiary parties, and the accounts to be debited and credited. SWIFT validates that the message is correctly formatted, routes it to the receiving institution, and keeps a record that it was sent. It does not hold an account for either party to the payment, and it does not clear or settle the instruction it carries.

A SWIFT code, more formally a Business Identifier Code, identifies an institution the way a postal code identifies an address. It says which bank the message is going to. It says nothing about where the money is, because at the moment a SWIFT message is sent, the money has not moved at all.

The message and the money travel separately

Our earlier post on how cross-border settlement works traced a payment through a chain of correspondent banks, each debiting and crediting nostro and vostro accounts on its own books. SWIFT is the layer that sits alongside that chain, not inside it. When an EMI instructs its correspondent to pay a beneficiary abroad, the SWIFT message and the settlement instruction typically travel together, hop by hop through the same chain of institutions, but they are two different things happening at each hop: a message confirming what should happen, and a book entry recording that it did.

This is why a SWIFT message can be delivered successfully while the underlying payment is still queued. The message arrives; the receiving institution's cut-off time, compliance screen, or funding position determines when the account entry that actually constitutes settlement gets made. SWIFT has no visibility into, and no control over, any of that. It confirms that an instruction was transmitted, not that value has changed hands.

What gpi added, and what it did not

SWIFT's global payments innovation service, launched in 2017, addressed a real problem with the messaging layer: a payment could disappear into the correspondent chain with no way for the originator to see where it was. gpi introduced the Unique End-to-End Transaction Reference, a 36-character identifier assigned when a payment is created and passed unchanged through every institution that handles it. Since November 2018, providing a UETR has been mandatory on in-scope SWIFT payment messages, and participating institutions confirm status updates against it in a shared tracker.

This materially improved visibility. A payment that once required a manual enquiry to each bank in the chain can now, in principle, be traced by a single reference. SWIFT's own reporting on gpi has cited a large share of payments confirmed as credited within around thirty minutes, and the great majority within a day, though these figures measure message-status confirmation and are not directly comparable to the beneficiary-leg timing the Financial Stability Board measures independently, which is a different question with a different answer.

What gpi did not do is shorten the chain, remove a cut-off time, or clear a compliance hold. A tracked payment sitting in manual review is exactly as delayed as an untracked one; the originator simply now knows it, in place of not knowing. Visibility is a genuine improvement over the alternative. It is not the same thing as speed.

The ISO 20022 migration: a bigger envelope, not a different road

SWIFT began migrating cross-border payment messages from the legacy MT format to the ISO 20022 standard in March 2023, running a coexistence period in which both formats were accepted. That period ended on 22 November 2025, a date SWIFT's board reaffirmed rather than moved, after earlier delays to the wider industry timeline. Since then, in-scope cross-border payment instructions must be exchanged as ISO 20022 messages, with legacy MT payment instructions no longer delivered on the network outside contingency processing.

The change is real and it is useful: ISO 20022 messages carry more structured data than MT ever could, including dedicated fields for addresses, purpose codes, and party identification, where MT often forced that information into unstructured free-text lines. Richer, better-structured data supports more automated reconciliation and reduces the false positives that trigger manual sanctions and AML review at each hop in the chain.

What it does not do is change where settlement happens or who performs it. ISO 20022 is a specification for the envelope, not a new road for the payment to travel. The correspondent chain, the nostro and vostro accounts, the cut-off times and the compliance queues described in our earlier post are all still there after the migration exactly as they were before it. An institution that expected the November 2025 deadline to solve a settlement-speed problem was solving the wrong problem.

Why the distinction is commercially real, not academic

  • Sanctions and network access are a SWIFT-layer question. Because SWIFT is a Belgian cooperative subject to EU and Belgian law, exclusion from the network is a messaging-access decision, imposed on the institution as a whole. It is a different kind of risk from a settlement delay on a single payment, and the two should not be diagnosed as the same problem.
  • Tracking is not control. A UETR lets an originating institution see that a payment is held at a particular hop. It does not let that institution accelerate release, waive a cut-off time, or clear a compliance hold sitting with an intermediary it has no direct relationship with.
  • The cost structure is unchanged. Because gpi and ISO 20022 operate at the messaging layer, they do not remove intermediaries from the chain. Lifting fees, FX spreads, and pre-funding requirements — the places cost accumulates that we set out previously — persist exactly as before.
  • Better data only helps if every hop uses it. ISO 20022's structured fields can reduce manual review, but only where the institution receiving the message actually validates against the structured data rather than falling back on the same free-text checks it always ran. Adoption of the format and adoption of what the format enables are not the same milestone.

Where the network model differs

SWIFT, gpi, and ISO 20022 all operate at the same layer: they improve how a payment instruction is described and tracked as it passes through a chain of correspondent institutions. None of them changes the chain itself.

UNYX, the cross-border settlement network for regulated financial institutions, operates at the layer underneath that. Members settle directly with each other rather than routing value through a chain of correspondent accounts, which is a structural difference from anything a messaging upgrade can produce. Only licensed financial institutions are admitted, and every member is licensed and KYC/AML checked before it can settle a transaction.

Where this is the wrong framing

Not every settlement problem is a messaging problem, and not every messaging problem is a settlement problem. An institution whose payments are slow because a correspondent's compliance queue is genuinely backed up has a different issue from one whose payments are slow because it has not yet completed its own ISO 20022 migration. And an institution concerned about SWIFT network access after a sanctions event is dealing with a regulatory question that a settlement network does not, on its own, resolve. Diagnosing which layer a delay actually sits at — the message, the compliance check, or the settlement chain itself — is the first step, and it is worth getting right before choosing where to spend the fix.

Frequently asked questions

Does SWIFT move money?
No. SWIFT operates a messaging network that carries standardised payment instructions between financial institutions. It does not hold accounts, hold funds, or perform clearing or settlement. The transfer of value happens through the correspondent accounts the institutions in the chain hold with one another.

What is SWIFT gpi?
SWIFT gpi is a service that tracks a payment message's progress through the correspondent chain using a fixed reference, the UETR, assigned when the payment is created. It improves visibility into where a payment instruction stands. It does not shorten the chain or remove the cut-off times and compliance checks that determine when the payment actually settles.

What is a UETR?
A UETR, or Unique End-to-End Transaction Reference, is a 36-character identifier assigned to a SWIFT payment when it is initiated and carried unchanged through every institution that handles it, allowing the payment to be traced by a single reference rather than a separate bank-assigned number at each hop.

What changed on 22 November 2025?
SWIFT ended the coexistence period between its legacy MT message format and the ISO 20022 standard for cross-border payment instructions. From that date, in-scope payments must be exchanged as ISO 20022 messages. This changed the structure and richness of the payment message. It did not change the correspondent banking chain the payment settles through.

Why doesn't tracking a payment make it arrive faster?
Tracking shows the status of the message, not control over the settlement it describes. A payment held at an intermediary's compliance queue or waiting for a local system to open is exactly as delayed whether or not the originator can see it happening.

Continue Reading

Read More