The affiliate never took a wager. The bank closed the account anyway.

An affiliate marketing an online casino has never held a stake. They do not take deposits, they do not hold player funds, and they are not named on the operator's gambling licence. They write content, buy media and send traffic. In most European markets they need no licence of their own, though Romania is an exception and licenses affiliates directly under a Class II authorisation. Outside Europe the picture differs again: New Jersey, Pennsylvania, Colorado, Indiana and Virginia all require affiliates to register as vendors or gaming service providers, with the tier and in some cases the requirement itself turning on how the affiliate is compensated. On any plain reading of what a bank is worried about, though, an affiliate is a marketing business.

Wise's published guidance disagrees, and it does so by name. The company's help centre states that it does not support "any business in any part of the world that's involved in gambling", and the examples it gives include "affiliate marketing for gambling businesses", with the warning that "we may close your business account as a result". That is a clear, honest and publicly documented position, and it is worth more to an affiliate than the silence of providers that may hold the same view without printing it. The uncomfortable part is what it implies about the rest of the market.

Why a bank treats a gambling affiliate as a gambling business

Start with the case for the bank, because it is stronger than the affiliate community usually allows. Licence condition 1.1.2 of the Gambling Commission's LCCP, "Responsibility for third parties", opens by making licensees responsible for the actions of third parties, and requires them to contract on terms that oblige those third parties to "conduct themselves in so far as they carry out activities on behalf of the licensee as if they were bound by the same licence conditions and subject to the same codes of practice as the licensee". On direct marketing specifically, the Commission's guidance on affiliates says that it and the Information Commissioner's Office "consider that you are primarily responsible for any breaches". Malta reaches a similar place from a narrower starting point: the MGA's commercial communications guidelines state, in the context of unsolicited communications and self-excluded players, that "both the affiliate and the authorised person may be held to be in breach of the Regulations".

Read that from a financial crime desk rather than a compliance desk. Affiliate conduct is regulated conduct. It attracts enforcement, some of it is attributable to a licensed operator, and the affiliate sits inside a value chain whose regulator has written down that responsibility flows upward. A bank is not making a category error when it scores that relationship as gambling exposure. It is reading the documents the regulator wrote and reaching a conclusion those documents invite.

Where the reasoning weakens is at the point of remedy. Refusing the relationship removes the exposure from the bank's book. It does not remove it from the system. The affiliate still gets paid — through a wallet, a network, a third-party payout provider or a stablecoin. Those are arrangements the bank cannot see, which is not at all the same thing as arrangements nobody can see.

Europe's supervisory record holds ten de-risking findings, and no gambling at all

The European Banking Authority's fifth biennial opinion on money laundering and terrorist financing risks affecting the EU's financial sector, EBA/Op/2025/10, was published on 28 July 2025. The report attached to it, EBA/REP/2025/22, carries a section on de-risking whose conclusion is a positive one, and it deserves to be stated fairly before any part of it is questioned.

The EBA defines de-risking as what a financial institution does, and treats it as unwarranted where "an institution does not take into account an individual customer's risk profile". It then notes that its 2024/2025 Consumer Trends Report "suggests that 'de-risking' remains an important issue for EU consumers", while recording that the information behind that report "does not distinguish between warranted and unwarranted de-risking". Turning to the supervisors, it finds that "40% of CAs indicate that unwarranted de-risking has decreased" and "another 40% suggest that de-risking is not an issue in their Member State", that "80% of CAs took action to tackle unwarranted de-risking", and that this assessment "is backed up by the small number of material weaknesses linked to de-risking in EuReCA (10 material weaknesses between 2022 and 2024)". The executive summary treats the trend as decreasing, while noting that the EBA "is now assessing possible next steps because access to basic financial products and services is an important public interest goal". The EBA is not contradicting itself here. It is explaining why two datasets disagree, and then choosing between them on the evidence it holds.

So take the number that should give an affiliate pause, which is ten. Ten material weaknesses linked to de-risking, recorded in the EBA's central database, across the whole of the European Union, over three years. There are three readings available. De-risking has become rare; or supervisors are examining it and judging most of what they find to be warranted; or the recording apparatus is not sitting where the practice happens. The document supports the first two and cannot speak to the third.

Which brings us to what is not in it. Across the sixty-nine pages of the opinion and its attached report there is no mention of gambling, gaming, betting, wagering, casinos or lotteries at all. Part of that is scope, and the objection should be made before anyone else makes it. The mandate runs to the EU financial sector. Gambling operators are obliged entities supervised elsewhere under Article 2(1)(3)(f) of the anti-money-laundering directive, and gambling affiliates — who wager nothing and take no stake — sit outside the Article 3(14) definition entirely. The absence is lawful and unsurprising.

It is also why ten is not the reassurance it looks like. That figure counts the occasions on which a supervisor examined a bank and judged its de-risking unwarranted. It is a measure of supervisory detection, not of how often accounts are refused — and nothing in either document suggests any supervisor has looked at this sector. Both of the channels feeding the picture point at consumers. The Consumer Trends Report aggregates national consumer protection authorities, consumer associations, industry associations and national ombudsmen. The EBA frames the public-interest goal as access to basic financial products and services. A business-to-business marketing company refused a business account has no obvious route into either. None of that makes the EBA wrong. It means this document cannot tell you whether it is right about gambling affiliates, and it should not be cited as though it could.

Two of the largest listed affiliates are moving in opposite directions, and both widen the payment surface

Better Collective and Catena Media both reported their second quarter this month. The temptation is to read them as one trend. They are not one trend, and the difference is the interesting part.

Better Collective's interim report of 20 August 2026 shows revenue of EUR 89.1 million, up 9%, with EBITDA before special items of EUR 27.0 million, up 20%. Recurring revenue — revenue share, subscription and CPM combined — was EUR 53.5 million, up 2%, with revenue share alone at EUR 43.6 million, up 5%. CPA revenue was EUR 19.5 million, up 11%. Sponsorship revenue was EUR 15.7 million, up 39%. In North America the company describes a deliberate transition into revenue share, and the figures bear it out: revenue share there grew 49% to EUR 6 million while CPA grew 50% to EUR 5 million. Those are, for practical purposes, the same number — which is worth knowing before anyone quotes the CPA figure on its own as evidence that the model is turning.

Catena Media's report of 11 August 2026 runs the other way. Revenue was EUR 9.5 million, down 1% year on year and down roughly 23% on the previous quarter, on our own calculation from the two reported figures. Sports revenue fell 43%. Revenue share fell from about 12% to about 6% of group revenue against the same quarter last year, while CPA rose from 86% to 88% and fixed fees from 2% to 6% — so most of what left revenue share went to fixed fees rather than to CPA. Direct costs rose from EUR 2.4 million to EUR 3.0 million, which the company attributes to increased activity in its MRKTPLAYS marketplace, now on the company's own account more than a third of group revenue. On the search environment the report is blunt: "changes in how users discover content mean that a given ranking position generates fewer clicks and hence less revenue than it did a year ago".

The commercial reading of those two reports is well covered elsewhere. The payments reading is not, and it does not require picking a winner. Revenue share is among the more bank-friendly payment profiles in this industry: few counterparties, monthly settlement, a familiar payer, a stable corridor. Most of the other lines are less so. Better Collective is now running revenue share, CPA, sponsorship and CPM as material lines at the same time, with its growth in a different currency under a different tax-reporting regime. Catena has added an entirely new counterparty type in a marketplace that is now a third of its revenue. Commercial diversification is, mechanically, payment diversification — more counterparties, more corridors, more instruments — and that second half of the sentence does not appear in either report.

The honest caveat: Better Collective's recurring revenue moved from 64% to 60% of the total, which is one quarter, and anyone reading a structural shift into it is overreading. The direction is clearer at Catena, where revenue share has roughly halved as a share of revenue. What is not in doubt at either company is the widening — and widening is what gets an account reviewed.

What an affiliate pays for having no bank

Affiliate programme terms are public documents and repay being read as payment instruments rather than marketing agreements. Betsson Group Affiliates' terms set a EUR 50 minimum payout threshold, with balances below it carrying into the next month, and separately a EUR 50 minimum withdrawal; commissions "calculated at the end of each month and payments shall be made on a monthly basis in arrears, not later than the 10th of the following calendar month"; payment made through the affiliate wallet "or such other means that we may determine from time to time at our sole discretion"; and the affiliate "solely responsible for all risk, costs and expenses" it incurs. Negative carry-over applies where a high roller generates negative commissionable revenue of at least EUR 10,000 in a calendar month.

Termination is automatic on any of four six-month triggers: fewer than ten new customers in total across six consecutive months, six months without accessing the affiliate account, six months without a withdrawal from the wallet, or six months of negative commission. Read the withdrawal trigger alongside the two EUR 50 floors and a small affiliate who never accrues enough to withdraw meets a termination condition by doing nothing wrong. These are not unusual terms, and by the standards of the sector they are unusually accessible ones — the new-customer limb is drafted as "failure to generate less than ten (10) New Customers", which says the opposite of what it means, but at least it is on a public page where it can be read at all.

Note what the discretion clause does. The method of payment is the operator's choice rather than the affiliate's, and the affiliate carries its own cost of performance. An affiliate who cannot hold a bank account has no leverage over that choice, so it takes the wallet — and the wallet prices the absence of an alternative. Skrill's published German schedule charges up to 3.99% on currency conversion, 2.99% with a EUR 0.50 minimum on sends to other Skrill accounts, reducible to 1.45% at the highest tier, and a EUR 5 monthly service fee once six months pass without a login or transaction. The Norwegian schedule carries the same six-month trigger. The US schedule sets it at twelve months and the fee at USD 5, which is worth knowing if you have landed on the wrong page. Conversion, transfer and dormancy are each charged, and each is charged to the party with the fewest options.

That cost does not stay with the affiliate. It arrives back at the operator as churn in the programme, as pressure on commission rates, and as an argument for whichever competing programme pays through a rail the affiliate can actually use.

What we do not know

Three things, and we would rather say so than write around them.

Whether this de-risking is wrong. The EBA does define the term: de-risking is unwarranted where an institution does not take into account an individual customer's risk profile. On its face a blanket category exclusion fails that test, and that reading cuts toward the affiliate. But a bank that genuinely cannot see an affiliate's downstream conduct, in a chain where the regulator has said responsibility flows upward to a licensed operator, may be making a defensible decision and explaining it badly. Whether a supervisor would call that unwarranted is a judgement made case by case, and we have not found one who has been asked about this sector. We are not persuaded either way.

We could not find the data. We looked for a survey of affiliate payment experience covering 2025 or 2026 — timeliness, cost, method, account closure rates — and did not find one. The public evidence is forum threads and a June 2026 trade report describing delayed affiliate payments ranging "from relatively small balances of around EUR 1,000 to outstanding debts exceeding EUR 40,000", with no named operators, no stated sample, and verification asserted but not shown. That is not enough to build an argument on, and we have not built one on it. It is a striking gap in an industry that measures everything else it does.

Whether paying affiliates in stablecoins is itself a regulated activity. Every national transitional permission under Article 143(3) of MiCA has now lapsed. Member states set their own end dates — several closed during 2025 — and the last possible one was 1 July 2026, so the question is no longer theoretical for anyone operating in the EU. Our reading is that paying your own affiliates in a crypto-asset is not obviously a crypto-asset service, since MiCA defines those as being provided to third parties, while sourcing, converting or off-ramping on an affiliate's behalf plainly is one. We have not found an authority directly on the point, and we would not want anyone relying on our reading of it.

What this means in practice

If you run an affiliate programme, the payout method is a term you hold at your sole discretion and have probably never priced. It is worth asking your finance team one question: how many of your affiliates are paid by a method they chose, and how many by the only method still open to them? The second number is your churn risk, and it is not in any dashboard you currently look at.

If you are an affiliate business, the question is concentration. Revenue diversification has had years of attention in this industry. Payment diversification has had almost none, and the single point of failure may not be a search algorithm — it may be one provider whose policy page can be rewritten without notice. As above, we could not find the data to size that, and we are not going to pretend otherwise.

We build on multiple licensed institutions offering the same service rather than on one, which is a position on that risk rather than a solution to it. It is also an argument against consolidating a payout programme with any single provider, ours included. If you read the stablecoin question differently, or if you have data on affiliate payment experience that we could not find, we would rather hear it than be right quietly.