Want to know how AI search is changing affiliate marketing?
A surprisingly simple challenge
AI search is upending how people discover information, how affiliate content is used and how their contributions are measured and rewarded.
During a recent APMA webinar, I spoke to two of the most experienced AI voices in the UK, James Bentley from Numerical and Alex Springer from OpenAttribution, about what this means for affiliate marketing. Here you’ll read about the main topics we covered and what the industry needs to do to tackle them.
Amid all the complexity, the challenge is surprisingly straightforward: how do we connect activity within AI search to business value and ensure the publishers producing the content are recognised and rewarded?
Channel challenges
AI-generated answers are becoming a much bigger part of search.
This presents a particular challenge to a channel reliant on clicks to track activity, measure performance and assign commission.
The industry therefore has the challenge of tracking the value delivered before and beyond the click while continuing to operate through its established performance method.
The Three Layers: Retrieval, Citation and Outcome
Kicking off with some of the basics, James Bentley introduced the ‘three-layer framework’, which provides a simple way to understand what happens to publisher content within AI search.
1. Retrieval
Retrieval happens when an AI crawler or agent accesses a publisher’s website and collects information from a page.
Server logs and specialist tools can show that a crawler associated with ChatGPT, Google or another AI service visited a particular URL, and this confirms that the content was accessed.
It does not tell us what the user asked, whether the content was used in the answer or whether it influenced a decision. Retrieval is therefore an important signal, but it is not evidence of an outcome by itself.
2. Citation
Citation happens when an AI-generated answer names or links to a source.
These links may appear in platforms such as ChatGPT, Perplexity or Google’s AI search products and if the user clicks a citation, that visit may appear in the publisher’s analytics or referral data.
Tools including Profound, Peec AI, Semrush and others can also test prompts and show which brands, websites and pages appear in AI-generated answers.
This provides a useful view of visibility and share of voice, but much of the data is based on prompts created or simulated by the measurement platform rather than accurately reflecting what people are actually searching for. It also isn’t, by itself, a reflection of the commercial value of the citation.
3. Outcome
The outcome is the conversion, lead, sale or other result.
This is the part that affiliate marketers know well. Networks and platforms can record a click and connect it to a transaction and some are also beginning to identify traffic arriving from AI platforms.
The problem is the space between citation and conversion. A user may see a publisher’s content represented in an AI answer, carry out further research and later buy through another route, but unless the user clicks the citation and remains within a trackable journey, much of that influence disappears.
Stitching the three layers together
In the webinar we talked about the challenge of connecting retrieval, citation and outcome data which are currently held in different places.
James gave an example where a publisher might have a page about Nike running shoes. At URL level, it may be possible to see:
- how often AI crawlers accessed the page;
- whether the page appeared as a citation in relevant AI answers;
- whether it featured anywhere within a tracked conversion journey;
- how users engaged with the page directly;
- whether sales of the products covered by the page changed.
Bringing those signals together would provide a more complete view of the page’s value.
It would not produce a perfect answer, but the affiliate channel has always operated from a position of not having every useful piece of data available.
Want to know more about the three layers? Download our AI report today.
Share of voice
In the webinar, we also discussed the importance of cross-industry collaboration needed to coalesce around new measurement standards, such as deciding how much weight to give each signal and how to separate correlation from incremental impact.
Share of voice is becoming one of the first widely used AI visibility measures.
This records how often a brand, product or publisher appears across a defined set of prompts compared with its competitors, which requires a leap of faith for affiliate marketers who are used to hard performance metrics.
However, AI search gives us a reason to look again at the early funnel influence and visibility that the industry has been trying to contextualise for many years. This is particularly true for content publishers whose role is frequently much earlier in the customer journey.
Share of voice is not likely to become a substitute for sales or ROI, but it could be an additional measure that helps explain how affiliate content contributes before a transaction takes place.
Commercial models
One other major area of discussion in the webinar was the commercial models that may emerge, by either necessity or innovation.
We have seen a shift away from last-click CPA and this year’s State of the Affiliate Nation report shows that around 20% of affiliate brand spend is accounted for by these payments.
This flexibility is a useful starting point because AI search will not be monetised through one standard model immediately. Citation and retrieval data could see fixed-fee campaigns, content partnerships or hybrid deals in which publishers are paid for both exposure and performance emerge. This will be a fast-developing story and, again, may need cross-industry support to establish standardisation.
Some technology providers are also beginning to explore whether crawl or citation activity can be connected to commissionable events.
Whatever the model, the same commercial question remains: what did the advertiser receive in return for their investment?
Both webinar guests agreed that it was imperative publishers and networks need to start developing their own remuneration models now rather than wait for LLMs to create their own monetisation platforms.
What Publishers can do
There are certain things cited in the webinar that publishers should be doing now, including:
- monitoring crawler activity;
- tracking citations and share of voice;
- identifying the pages most frequently used by AI platforms;
- combining this with audience and engagement data;
- including relevant AI visibility measures in advertiser proposals.
Publishers should also understand which subjects, formats and pages attract AI retrieval and citation. That information can inform both editorial planning and commercial discussions. APMA data shows that only one in five UK affiliates are actively measuring AI visibility and monitoring or managing AI crawlers. Many are simply working blind, which absolutely needs to change and fast.
What Brands can do
James also touched on what brands should be doing, namely asking their networks, agencies and technology providers what they can already measure.
They should compare those capabilities against the three layers:
- Can they see retrieval?
- Can they measure citations?
- Can they identify traffic and conversions from AI platforms?
- Can those datasets be connected?
- Where are the gaps?
Brands should also look at the information they make available to AI systems. Accurate product feeds, structured information and clear content all affect whether products can be understood and recommended.
This will become particularly important as AI moves from answering questions to assisting with purchases.
What Networks and Agencies can do
Networks and agencies are best placed to help connect the different datasets.
They already sit between advertisers and publishers and hold much of the click and conversion data. By adding citation and retrieval information, they can help create a broader view of the customer journey. In the short term, this may involve manual analysis, third-party tools and individual campaign arrangements.
Networks and agencies can also help establish consistent terminology and avoid each provider presenting fundamentally different measures under the same label. Alex Springer made the point that while product roadmaps are important, now is the time for senior stakeholders across the industry to come together to collaborate on the bigger picture.
Agentic shopping
We also touched on agentic shopping whereby an AI agent may research products, compare options and prices, then add an item to a basket and complete a purchase.
It could also monitor prices and act when a product reaches a specified level.
This compresses discovery, consideration and purchase into a single process, completely collapsing the sales funnel.
Publisher and affiliate identifiers need to survive when a journey moves from an AI assistant to a retailer’s website, app or checkout system. If they do not, the content or partner responsible for the recommendation may receive no credit. While detail is lacking at the moment, networks, advertisers and commerce platforms should start testing these journeys before agent-led purchasing becomes more established.
Regulation and collaboration
Finally, we touched on how regulators are beginning to examine how AI search platforms use publisher content and how fairly they deal with businesses that depend on search.
The CMA’s work on Google’s search services and the implementation of the EU AI Act will influence transparency, accountability and access to data.
Beyond that, affiliate marketing also needs more collaboration of its own. The affiliate industry still relies heavily on individual platforms, methodologies and commercial definitions which have created a highly fragmented ecosystem..
A common approach to retrieval, citation and outcome data would make it easier for publishers, brands, networks and agencies to compare results and agree how value should be recognised.
