James Bentley, the author of our AI guide ‘Navigating The New Rules of Discovery: How AI is Reshaping Affiliate Marketing’ hosted a members only Q&A session today.
Following the publication of the APMA’s The New Rules of Discovery: How AI is Reshaping Affiliate Marketing, we invited the report’s author James Bentley, founder of Numerical and Chair of the APMA AI Task Force, to an open Q&A.
Here are some of the biggest questions we tackled.
Can affiliate attribution survive agentic commerce?
This is perhaps one of the biggest questions hanging over the industry.
Today, a consumer might research a product, visit a publisher, click through to a retailer and purchase. But agentic commerce could fundamentally change that journey.
A consumer could instead tell an AI agent what they want, give it certain parameters – including price, preferred retailers or rewards – and ultimately delegate some or all of the purchasing decision to the agent.
So where does affiliate attribution fit? James believes there are reasons to be optimistic.
There are already early examples demonstrating that affiliate recognition can technically be incorporated into agentic transactions. We’ve also seen demonstrations of future shopping experiences incorporating third-party loyalty and points providers.
The technology therefore doesn’t inherently remove affiliates from the equation.
The bigger issue is whether the affiliate industry can ensure its attribution mechanisms become part of the infrastructure being built.
The takeaway: affiliate attribution can have a place in agentic commerce, but the industry needs to engage early. If the standards and infrastructure are built without us, retrofitting affiliate attribution afterwards will be considerably harder.
If an AI agent automatically finds cashback, does cashback still add value?
This prompted one of the most interesting debates of the session.
Imagine a consumer tells their agent to buy a product – and the agent automatically routes the transaction through a cashback provider.
The cashback publisher hasn’t necessarily influenced the original product decision. So should it still receive commission?
James’s view is that we may need to think about value differently.
A consumer could explicitly tell their agent to only purchase from a retailer offering cashback. In that situation, cashback has become part of the consumer’s declared purchasing criteria.
And the cashback provider is still delivering infrastructure, processing payments and providing the financial reward that influences where the transaction takes place.
That doesn’t mean questions around incrementality disappear. Brands may eventually need to conduct incrementality studies on agent behaviour, just as they do on human customer journeys today.
The takeaway: agents won’t necessarily make cashback irrelevant. They could actually make rewards more systematically incorporated into purchasing decisions – although that will inevitably reignite the industry’s debate around incrementality and value.
Who are likely to be the winners and losers from AI search?
There isn’t a simple answer, but some affiliate models appear more insulated than others.
James sees closed-user groups, loyalty and cashback businesses as relatively well protected because their proposition relies on membership, logged-in experiences, rewards and identifiable transactions.
Their challenge could increasingly become discovery: how does an AI agent discover the platform and know that a consumer is eligible for its rewards?
At the other end of the spectrum, traditional content publishers face greater uncertainty.
Product reviews, comparisons and “best X” articles have historically performed well within search. AI can increasingly synthesise that information itself – potentially reducing the need for a user to visit the publisher.
There are also signs that AI search platforms can change the types of sources they favour extremely quickly, including potentially prioritising official brand sources over third-party comparison content.
Voucher publishers face their own challenges, but there could also be an interesting opportunity around unique codes. If an AI surfaces a publisher’s exclusive code and that code is redeemed at checkout, it creates a clear mechanism for attributing value – even if the consumer never visits the publisher’s website.
The takeaway: the affiliate models most reliant on generating a traditional click from search potentially have the greatest exposure. Models with another identifiable mechanism for proving value – membership, cashback, loyalty, unique codes or proprietary data – may be better protected.
Is AI really going to have that much impact on ecommerce?
Right now, many businesses still aren’t seeing huge volumes of identifiable ecommerce traffic arriving from AI platforms.
James’s answer to whether that means the impact is being overstated was emphatic: no.
One of the problems is that we’re measuring what we can see.
A consumer could conduct multiple product searches, comparisons and follow-up questions inside an AI platform before clicking anywhere. None of that research is necessarily visible to the publisher, brand or affiliate network.
Meanwhile, the major AI and search platforms are investing heavily in shopping functionality, product information, advertising and agentic experiences.
The direction of travel is towards AI playing a much bigger role in product discovery and purchasing – even if the final model is still evolving.
The takeaway: don’t use today’s visible AI referral traffic as a proxy for AI’s actual influence on ecommerce. A significant proportion of that influence may already be happening before the first measurable click.
Could product data become one of the affiliate industry’s biggest assets?
If AI agents are going to help consumers choose products, they need accurate information.
Not just a product name and price, but potentially specifications, sizes, colours, availability, compatibility, safety information, manuals, reviews and much more.
That creates an interesting opportunity for businesses that already specialise in gathering and structuring product information.
James described this as the potential for “information as a service.”
Comparison businesses, CSS providers and other organisations within the affiliate ecosystem already have significant expertise in extracting, organising and maintaining product data.
In an AI-first discovery environment, the value of that capability could increase considerably.
The takeaway: some businesses may need to stop thinking purely about how they get traffic from AI platforms and start considering whether the information and data they hold could become valuable to AI platforms.
Could people manipulate what AI platforms recommend?
James preferred the word “influence” – but either way, the answer is yes. And it’s already beginning.
New technologies can identify when an AI agent arrives on a webpage and potentially present information intended specifically for that agent.
That opens the door to a new form of machine or agent-oriented advertising: rather than advertising directly to the consumer, you’re attempting to influence the AI system gathering information on their behalf.
But there’s an obvious tension.
AI providers want their assistants to provide useful and trustworthy answers. They aren’t necessarily going to welcome third parties attempting to influence those answers outside of the platforms’ own advertising ecosystems.
We should therefore expect a familiar cat-and-mouse game to emerge, much as we have seen throughout the history of search.
The takeaway: influencing AI recommendations could become a significant new marketing discipline, but questions around disclosure, transparency and acceptable practice will become increasingly important alongside it.
Could an AI platform simply become an affiliate itself?
Another question from the audience went straight to the potential commercial conflict at the heart of all this.
If an AI platform influences the decision and facilitates the transaction, what stops it from becoming a publisher itself and taking commission?
James thinks it’s more likely that the major AI platforms will develop their own marketplaces, advertising and attribution ecosystems, rather than simply operating like a conventional affiliate publisher.
We’ve seen similar models emerge from major social platforms.
And as AI companies build advertising businesses alongside shopping functionality, we shouldn’t be surprised if they increasingly attempt to own more of the commercial infrastructure surrounding those transactions.
The takeaway: the bigger competitive threat may not be an AI platform joining an affiliate network. It could be AI platforms building alternative ecosystems of their own.
Why did OpenAI scale back native checkout – and will it come back?
Agentic purchasing sounds simple in theory.
In reality, buying the correct product, in the correct size, colour and specification, at the right price, from the right retailer is extremely complicated.
James believes early attempts at native AI checkout underestimated that complexity.
Agents need access to extremely rich and reliable product information, while integrations with retailer checkout systems introduce another layer of difficulty. Add the risk of an AI agent simply purchasing the wrong thing and it’s easy to see why progress hasn’t been straightforward.
But that doesn’t mean agentic shopping has disappeared.
James expects a more gradual evolution. One of the most obvious early use cases could be delegated price watching: tell an agent which product you want and the price you’re prepared to pay, then allow it to purchase when those conditions are met.
The takeaway: agentic checkout may have moved more slowly than initially expected, but the underlying direction hasn’t changed. Expect it to return as the technology, product data and consumer behaviour mature.
Could Black Friday accelerate all of this?
This could be one of the most important questions for the remainder of the year.
Compared with last Black Friday, AI search is available to significantly more consumers and increasingly embedded within mainstream search and shopping behaviour.
Does that mean we’re about to see a dramatic impact on affiliate performance?
James was cautious about making a firm prediction, but believes we should expect AI to play a much more prominent role.
And that raises an important measurement problem.
If consumers increasingly use AI to research products before eventually clicking or purchasing elsewhere, businesses could see changes in their conventional performance data without being able to clearly identify AI as the cause.
The priority ahead of peak should therefore be measurement.
What technology do you have in place? What AI activity can you see? And when you review peak performance afterwards, will you have enough information to understand the role AI may have played?
The takeaway: Black Friday could provide one of the clearest tests yet of AI’s impact on affiliate ecommerce. Brands and publishers should decide what they want to measure before peak, rather than trying to reconstruct the picture afterwards.
What should the affiliate industry be doing next?
There is still a huge amount we don’t know and that’s precisely why the APMA’s work in this area will continue.
The AI Taskforce is already looking at an update to The New Rules of Discovery, including potentially expanding the measurement framework introduced in the original report.
Our next major focus will be AI and content – looking at how publishers are using AI to create content, what good practice looks like and where the industry may need greater transparency or guidance.
We’re also beginning to explore agentic readiness: the practical steps brands and publishers can take to make their information more accessible and useful to AI agents.
Because while nobody can predict exactly what AI-driven affiliate marketing will look like in two or three years’ time, one thing is increasingly difficult to argue with:
The customer journey is changing. The affiliate industry needs to make sure its measurement, commercial models and standards change with it.
The APMA’s three-part report, The New Rules of Discovery: How AI is Reshaping Affiliate Marketing, explores AI search, measurement, attribution and emerging commercial models in more detail.
Members can catch up on the webinar here
