Retail media built a perfect advertising machine. AI has disrupted how it’ll work
Opinion – Week in Focus
A category that cannot yet agree on how to measure what it does is poorly placed to absorb a fundamental change in how the customer journey works. It’s an issue that retail media will need to address, and quickly.
Retail and commerce media have had an extraordinary run. Global spend hit approximately $140bn in 2024 (EMarketer) and is forecast to exceed $200bn by 2026 (WARC).
In the United States alone, the market grew 20% year on year in 2025 (Nielsen). Europe accelerated at 22% – more than three times the rate of total ad spend growth (Adtelligent).
Brands have poured budget in because the logic was sound: meet customers where they shop, close the loop between ad exposure and purchase, and do it all with first-party data that actually works. For several consecutive years, retail media was the easiest growth story in advertising.
However, recent consumer data suggests that the story is about to get considerably more complicated.
Kantar surveyed 2,000+ UK consumers in January and found that one in three would now complete a purchase directly through an AI tool like ChatGPT, bypassing retailer websites entirely.
One in four already uses AI for product recommendations. Travel and tourism led all categories as the sector where consumers are most likely to consult AI first, ahead of grocery, fashion, and telecoms.
These findings describe something more significant than a shift in consumer behaviour. They describe a structural challenge to the customer journey on which the retail media model was built.
The infrastructure problem hiding in plain sight
Retail media’s foundational promise has always been proximity to purchase.
Sponsored placements, onsite search, display units, loyalty data – all of it sits inside a walled environment where the retailer controls the context, the data, and the conversion point. The whole architecture assumes the customer arrives. They browse. They are addressable.
An AI-intermediated buying journey breaks that assumption before it starts.
When a consumer asks an AI model to recommend a travel product or a retail category, they are not browsing a website. They are receiving a curated, context-rich shortlist from something they experience as authoritative and, increasingly, sufficient.
The consideration phase happens inside the LLM interface. The decision may follow immediately after. The retailer’s owned environment – the digital real estate on which billions of pounds of ad spend is predicated – is bypassed entirely, at least until the point of transaction.
The brands that appear in that AI-generated summary are there not because they won an auction, but because the model had enough coherent, consistent, credible information about them to place them with confidence. The brands that don’t appear have lost the consideration phase without ever knowing they were in it.
A category scaling into headwinds
The timing matters. Retail media has grown at speed, but has done so whilst carrying unresolved structural problems.
According to IAB Europe’s Attitudes to Retail Media Report 2025, 78% of brands and agencies identify measurement as the area most in need of standardisation.
Fewer than one in six retail media advertisers report strong confidence in their measurement of effectiveness, and nearly half admit they are struggling (Skai).
Getting the industry to agree on definitions for even baseline metrics took 18 months of negotiation, and adoption remains uneven (IAB/MRC, January 2024).
A category that cannot yet agree on how to measure what it does is poorly placed to absorb a fundamental change in how the customer journey works. The AI disruption does not arrive in a stable, well-governed ecosystem. It arrives in a fragmented one.
Where the power is shifting
Retail media’s growth has been driven by three things: attributable ROI, first-party data, and closed-loop attribution. All three are weakened, to varying degrees, in a world where AI mediates discovery and consideration before a customer ever reaches the owned environment.
Closed-loop attribution requires the loop to close inside the retailer’s estate. First-party data requires the customer to show up. ROI measurement requires the touchpoints to be visible and ownable.
AI platforms have something the retail media networks don’t: the full picture of consumer intent before it crystallises into a visit, a search, or a click. Kantar’s data shows 41% of UK consumers are already fully open to acting on AI recommendations, with a further 19% moving in the same direction. The audience and the data are beginning to live somewhere else.
The question retail media has not yet asked loudly enough is whether networks built on owned website infrastructure can evolve fast enough to stay relevant in a buying environment that may increasingly form its decisions before the website is ever opened.
Measurement standards, format consistency, and closed-loop data are already unsolved problems. AI intermediation introduces a challenge that lies entirely outside the existing toolkit.
The growth was real. The machine worked. It was just built for a customer journey that is now being redrawn.
Paul Evans is the founder and chief positioning engineer at V2RSION