Affiliate Marketing Wasn’t Built for AI Agents — Here’s Why That Matters

Affiliate marketing is a $17 billion global industry. It powers an estimated 16% of all e-commerce sales in the United States alone. Every major retailer, SaaS company, and service provider runs some form of affiliate or partner programme. It works.

Until it doesn’t.

The entire system — every tracking pixel, every cookie, every last-click attribution model — was designed around one assumption: a human being is using a web browser to shop. That assumption held for twenty years. It’s breaking now.

The numbers that should worry the industry

AI-referred traffic to websites grew 527% between January and May 2025, according to data from SparkToro. ChatGPT now has over 900 million weekly active users. Perplexity processes more than 500 million queries per month. Google’s AI Overviews reach 1.5 billion users across 200+ countries.

Gartner projects that traditional search engine traffic will decline by 50% by 2028 as AI-powered discovery replaces the search-click-browse pattern that affiliate marketing depends on.

This isn’t a future prediction. It’s already happening. AI agents are recommending products, comparing prices, and guiding purchase decisions — and the affiliate infrastructure has no way to track any of it.

How affiliate tracking actually works (and where it breaks)

To understand why AI agents break affiliate marketing, you need to understand how the tracking chain works today.

Step 1: The click

A publisher — a blog, a comparison site, an influencer — places a tracked link on their content. When a human clicks that link, the affiliate network (Awin, CJ, ShareASale, Impact) records the click and drops a cookie in the user’s browser.

Where it breaks: AI agents don’t click links in a browser. When ChatGPT recommends a product, there is no click event. When a Perplexity answer mentions a service, the user might go directly to the merchant’s site. The tracked link is never touched.

Step 2: The cookie

The cookie stored in the browser contains the affiliate’s identifier and a timestamp. It typically lasts 30 days. If the user returns to the merchant’s site within that window and makes a purchase, the affiliate gets credit.

Where it breaks: AI agents don’t have browser cookies. Even when they do operate a browser instance (like ChatGPT’s browsing mode), the session is ephemeral. Cookies are not persisted between sessions. The entire cookie-based attribution model is invisible to AI-driven commerce.

Step 3: The conversion

When a purchase happens, the merchant’s site fires a tracking pixel or server-side postback to the affiliate network, matching the conversion to the stored cookie. The affiliate is credited and paid.

Where it breaks: If there’s no cookie, there’s no match. The conversion happens, the merchant makes a sale, but the system has no record of what influenced the purchase. The agent that drove the recommendation gets nothing. The merchant has no idea which agents are sending them business.

This isn’t a technical glitch — it’s a structural failure

The problem isn’t that affiliate networks need to update their JavaScript. The problem is that the fundamental architecture — click → cookie → convert → attribute — doesn’t map to how AI agents operate.

AI agents work through a different pattern entirely:

Discovery: The agent queries a structured data source or crawls merchant information
Evaluation: The agent compares options based on user preferences, price, reviews, and availability
Recommendation: The agent presents one or more options to the user with reasoning
Purchase: The user may buy directly, visit the merchant site, or ask the agent to complete the transaction

At no point in this flow does a traditional affiliate link get clicked. At no point does a cookie get set. The entire attribution chain is absent.

What the industry is doing about it (not much)

Most affiliate networks are focused on two things right now: influencer marketing and retail media. Both are browser-based, human-driven channels. They’re growing, but they’re not where commerce is heading.

A few networks have started experimenting with server-to-server tracking and first-party data models, which reduce cookie dependency. But these solutions still assume a human-initiated browser session. They don’t account for agent-mediated commerce at all.

The industry trade bodies — the Performance Marketing Association, the Internet Advertising Bureau — haven’t published guidance on AI agent attribution. There are no standards. There are no working groups. The conversation hasn’t started.

Meanwhile, AI traffic converts at 4.4x the rate of traditional organic traffic, according to industry data. The highest-quality traffic source in digital commerce is the one nobody can attribute.

What needs to change

The affiliate industry needs three things to survive the agent era:

1. Attribution that works at the protocol layer

Attribution can’t depend on browser-level mechanics (cookies, pixels, JavaScript) when the intermediary isn’t a browser. It needs to work at the data layer — embedded in the transaction itself, not in the user’s session.

2. Structured offer data that agents can consume

Today, agents that want to recommend products have to scrape merchant websites, parse unstructured HTML, and hope the data is accurate. This is brittle, expensive, and often blocked by anti-bot measures. Merchants need a way to publish offers in a format that agents can reliably read and compare.

3. A commission model that rewards recommendations, not clicks

The last-click model made sense when the last click was the point of influence. In agent-mediated commerce, the point of influence is the recommendation — which might happen in a conversation, a comparison table, or a structured API response. Attribution needs to follow the recommendation, not the click.

The window is closing

Brand mentions are now 3x more strongly correlated with AI visibility than backlinks, according to Ahrefs research from December 2025. Only 23% of marketers are investing in any form of AI search optimisation. The gap between where commerce is going and where the infrastructure is today is widening every quarter.

The companies that build the attribution layer for agent-driven commerce will define the next era of affiliate marketing. The ones that wait for the industry to agree on a standard will find that someone else already built it.

This is the first in a series on agent-native commerce. Next: what happens when AI does the shopping.