Up to 70% of AI-referred retail traffic arrives without a referral tag and is counted as direct. Your budget decisions are being made on incomplete data. Here's what's really happening.
Key takeaways
There's a number sitting in your retail media analytics right now that looks fine. Stable. Maybe even healthy. And it's quietly lying to you.
Your direct traffic figure.
Here's what's actually happening: when a shopper uses an AI assistant (ChatGPT, Gemini, Perplexity, Alexa) to research a product and gets a recommendation, they often follow that recommendation directly to a Product Detail Page. No search query on the retailer's site. No banner click. No referral tag.
Just a landing. Which your analytics platform counts as direct traffic.
Kiri Masters, host of Retail Media Breakfast Club, put a number on this at Signal to Scale 2026: up to 70% of AI-referred traffic arrives without a referral tag and is being systematically misattributed by retailers. The influence of AI on your funnel is not a future consideration. It's happening now, at scale, and most brands are measuring it as something it isn't.
Why This Matters More Than It Sounds
Attribution errors are easy to dismiss as a data quality problem. They're not. They're a budget allocation problem.
If 70% of your AI-referred traffic is being counted as direct, you're under-crediting the upper-funnel and off-site tactics that influenced those shoppers before they landed. You're over-attributing last-click placements. And you're making investment decisions, where to put more budget, which channels to scale, which to cut, based on a picture of the funnel that's missing a growing and increasingly important slice.
Over time, that compounds. You optimize harder toward last-click signals. Upper-funnel investment looks inefficient by comparison. You cut it. The pipeline of AI-influenced shoppers who would have found their way to your PDP quietly shrinks, and you don't see it coming because your direct traffic number still looks fine.
The 'Single-SKU Shopping Mission' Is the Mechanism
Kiri gives this behavior a name: the single-SKU shopping mission. A shopper opens their AI assistant, asks a specific question ('what's the best protein powder for endurance athletes under $40'), gets a specific answer, and arrives on a PDP already decided. They didn't browse your homepage. They didn't use your search bar. They didn't click your sponsored placement.
They were pre-sold before they reached you. And your attribution model doesn't know they ever consulted AI.
The implication for sponsored search is significant: your placements are being bypassed by a growing cohort of high-intent, ready-to-buy shoppers. Not because your ads are poor. Because the discovery moment happened somewhere else.
How to Start Closing the Gap
The solution isn't to wait for platforms to fix their attribution logic. It's to get a view of the consumer journey that's complete enough that referral gaps don't drive your strategy.
Amazon Marketing Cloud is the most powerful tool available for this. It connects impressions, search queries, and purchase events across the full Amazon funnel including signals that Sponsored Ads reporting never surfaces. With AMC, you can start to see which upper-funnel touchpoints are actually influencing the shoppers landing on your PDPs, regardless of what the referral tag says.
Xnurta AMC Hub puts that capability in front of your whole team. Pre-built analytics models. Custom audience creation. No SQL required.
It is an analytical approach requiring AI to show its work, providing full transparency into the data retrieved, the entities resolved, and the logical evidence chain behind every recommendation, ensuring every number is traceable.
Generalist models often sound authoritative while using incorrect data. In retail media, a recommendation built on wrong numbers leads to wasted spend and faulty budget reallocation, making accuracy the only true performance metric for your stack.
Evaluate tools using a benchmark framework that prioritizes retrieval accuracy, entity resolution, and visible reasoning rather than just writing quality. A trustworthy tool must allow you to verify the logic and data behind every specific recommendation it makes.
No fluff. Just what's working.
