The gaming affiliate industry has a problem.

The problem with building on someone else’s platform is the foundations are unstable, and for affiliates the changes have rarely been beneficial. AI summaries now answer a growing share of those queries directly (with interesting implications for who the publisher is). Google spent twenty years training people to type commercial questions into a box, and is now answering them itself. The regulators keep tightening their rules as well. Mostly this is to combat the egregious end of the affiliate spectrum - the untargeted bonus-spam pages, the opaque traffic - but the rules being written tend to have a disproportionate impact on the companies that adhere to them. Underneath both is a commoditisation of the offering. When your product is a ranked table of other people’s offers, there are only so many ways to arrange it. Every list looks like every other list, and every ‘expert review’ reads like every other expert review.

This is not a cyclical issue but a long term transition. I wrote in The legendary poker boom about markets that drain a reservoir once and mistake it for a repeatable trick. Intent capture has the same shape. The demand it harvests is finite in any given moment, and the mechanism that delivered it - a landing page, an SEO ranking and a human clicking on a link - is clearly in decline. The industry knows this, and is trying to find different models to replace that revenue flow.

Fundamentally there have only ever been two ways to make money connecting gaming operators to customers. You can harvest intent at the specific point it appears, or you can be the general substrate within which those sorts of decisions are made. Take the attention of an audience and circumvent the SEO bun fight. Even better if you can turn that attention into intent that wasn’t there in the morning. Better Collective, the largest of the listed affiliates, runs both models successfully. Action Network is the harvest: an intent asset waiting at the end of a search query. Action Podcasts is the substrate: hours of listening time from an audience who came for the content, with offers dropped into the middle of it. One designed for the point after the moment, the other designed to be there before the moment. This is the attention model.

You can see which model has the future by watching where the money is moving. Better Collective’s Paid Media segment - paid acquisition routed into BC assets and then operators, the attention model run with media buying instead of SEO rankings - is the fastest-growing part of the business, running at around a 25% margin, and they are still ramping the spend. Legend, probably the most successful affiliate with owned audience relationships, ran at around a 50% margin - a level the listed harvest businesses have never approached - and was bought by Genius Sports. Mark Locke’s defence of that deal is really all about this difference - he refers to a technology company, with an audience and the value of a participation layer. The market didn’t agree, but I think the rationale is correct.

The difference between the two models is what happens to the customer at the moment of monetisation. An intent click is consumed. It is spent, once, and nothing survives the transaction - no relationship, limited data, rarely a second act. The affiliate might never see that customer again, or might not know if it was an old customer who had returned.

An engagement product is interacted with. The customer comes back, and every return provides more data and more opportunities to surface relevant offers. You learn who they are and what they respond to from real behaviour rather than just a keyword. You can time offers to a lifecycle instead of a landing. For a customer who supports a certain team, this might look like one thing during the season, another after it, something different again when their team sacks the manager on a Tuesday. You can experiment on an audience that keeps showing up. And you can also do the one thing a ranked list will never do, which is make someone feel something. Offers land very differently on a person who is feeling something.

Intent or attention. Almost everything else in the comparison - the data, the targeting, the insulation from platform whims, the regulatory fit - falls out of that one distinction.

The operators have run attention products for decades - play-money poker tables, free-play casino teasers, the bracket games and predictors bolted onto every sportsbook. At PokerStars the play-money game was one of the most important funnels the company had. It allowed us to advertise and onboard customers, and I spent a chunk of my career on the machinery that turned free players into depositors. The model worked well, although the product was not ideal for purpose.

But the operator version is captive. It exists to feed one funnel, often one brand, and lives inside the regulated perimeter, where every feature must survive compliance and every message must align with the main product. The other version of the attention product went the opposite way. Zynga took the audience and walled it off, monetising through in-app purchases while the gaming operators pay Meta for the privilege of advertising at Zynga’s players - buying back their own prospects at retail.

Which leaves a third thing, and it has existed in various guises over the years, but mostly in fairly unsophisticated forms: the attention product run from the affiliate layer. Free in precisely the dimensions the operator version is constrained. Free to build features no regulated operator could ship. Free to market in a tone no compliance department would sign off. Free to monetise across many operators and several verticals, choosing the offer per customer, per moment, rather than funnelling everyone toward one brand’s welcome bonus.

The industry is moving fairly quickly in this direction. Better Collective is building a range of attention assets that fit the paid media model. Podcasts and forums have existed for a long time, but follow the logic and one underdeveloped endpoint is games. Not social casinos, which monetise by never letting the audience leave. Not operator free-play teasers, which are a vertical in fancy dress. Purpose-built games - entertainment products where the audience is the asset and the funnel is the business model, horizontal by design. Legend seems to see the same opportunity - it has been hiring social game developers. Job adverts are the most honest strategy documents most companies publish.

A good game ticks every box the intent model is losing, and at its best it has preferable attributes to the other attention options. It owns its audience instead of renting a position on a results page, and gets touchpoints in the dozens rather than a single click. It produces first-party data from the first session and emotion by the second. The iGaming cohort should filter itself, because the message, and the game itself, can be designed to target the right people. No algorithm change can take any of it away, and the whole thing is shaped like the direction regulation is heading - known customers, positive experiences and logged offers - rather than straining against it.

So why has nobody done this so far? Because it is extraordinarily hard. A landing page can be commissioned. A podcast can be recorded. A game people return to voluntarily cannot just be wished into existence. It needs a product good enough to compete with actual entertainment for actual attention, welded to affiliate machinery, marketing production and compliance infrastructure - the skills of three different companies in one product, where the failure of any one kills the others. And the companies best placed on paper are the least able in practice. The operators cannot ship the features that make it work. The social studios have no reason to send their players anywhere. Affiliates are not game studios, and if it were simply a question of hiring developers then great games would be easy.

For twenty years the job of an affiliate was to appear at the top of a web page. The next ten belong to the ones who can build something people come back to.

Philip Atkinson, CEO, July 2026