Retail Media's AI Turn

There's a version of the retail media story that gets told as a triumph of first-party data. Retailers have the most valuable purchase intent signals in the world — they know what people actually buy, not just what they browse — and they've figured out how to monetise those signals as a high-margin advertising business. Amazon built this playbook and the rest of the industry has spent several years reverse-engineering it.

The numbers are now big enough that "retail media" is no longer a niche term. Global retail media ad spend reached an estimated $122 billion in 2024 according to eMarketer projections, a figure other research firms place somewhat lower depending on how they define the category. Amazon Ads commands roughly three-quarters of the US retail media market by most analyst estimates. Walmart Connect, the second-largest US player, grew 24% in its most recent quarter and is scaling at a rate that makes it one of the fastest-growing advertising businesses globally. Beyond the US giants, UK grocers (Tesco, Sainsbury's, and Boots) have each built out media network offerings that are growing quietly and considerably.

All of this is well-documented. What's less well-covered is what happens when AI enters the picture.

What AI Is Actually Doing in Retail Media

Amazon has been first and most aggressive. Through 2024, it rolled out a suite of AI-powered campaign tools that can generate ad creative, build audience targeting strategies, write Amazon Marketing Cloud queries using natural language, and optimise bids in real time across its advertising inventory.

The pitch to advertisers is speed and performance. Instead of spending hours building audience segments and writing creative variants, a campaign manager can describe what they want in plain language and have the system generate it. Amazon's own marketing materials point to conversion lifts from AI-optimised targeting versus manually-built campaigns, though the underlying methodology for those comparisons isn't disclosed.

The pitch is real. So are the questions it raises.

The Transparency Problem

When a campaign manager builds a targeting strategy manually, they know exactly what signals they've used and which audiences they've included. When an AI system builds it, that visibility goes away. You see the inputs and the outputs; what happens in between is opaque.

This matters for several reasons. It matters for brand safety: you'd like to know your products aren't being advertised to audiences in contextually inappropriate settings. It matters for compliance, particularly as algorithmic targeting comes under increasing regulatory scrutiny. And it matters for the basic principle that advertisers should understand where their money is going.

There's also a more structural concern. Retail media AI learns to optimise toward conversion. But conversion within a closed retail ecosystem is a narrower objective than building a brand or growing a customer base. A system optimising for Amazon conversions isn't necessarily optimising for your brand's long-term health, and those objectives can diverge considerably.

See also: Retailer AI Backlash, which covers how some brands are starting to push back on the opacity of AI-driven retail decisions.

Who Owns the Customer Relationship?

Retail media at scale raises a question that brands have been quietly anxious about for a while. When a customer buys your product through Amazon after seeing an Amazon-served ad targeted using Amazon's first-party data, optimised by Amazon's AI, and fulfilled by Amazon's logistics, where exactly does your relationship with that customer sit?

The honest answer is: it's thin. Amazon knows who bought. You know you sold. The customer relationship, the data, the future marketing opportunity: these sit with the platform.

For brands with strong direct-to-consumer channels, this is a reason to think carefully about how much of your advertising budget flows through retail media versus owned channels. For brands without those channels, the calculus is harder. The sales are real. The customer relationship is attenuated.

The UK Opportunity (and Risk)

UK grocery retail media is worth watching specifically. Tesco's Clubcard dataset represents one of the most comprehensive purchase-behaviour records in the country, built over decades of loyalty data across a significant proportion of UK households. The Tesco Media and Insight Platform, operated with dunnhumby, gives brand advertisers access to targeting built on that data. Boots' Advantage Card similarly covers health and beauty purchasing in a category where purchase intent signals are particularly sensitive and commercially valuable.

Neither Tesco nor Boots has moved as fast as Amazon in rolling out AI-powered targeting tools for their media networks, but the trajectory points that way. When they do, the same transparency and attribution questions will apply. The UK's regulatory environment, including the CMA's ongoing attention to algorithmic pricing and platform practices, means the compliance dimension will be sharper here than in the US.

What Advertisers Should Be Asking

The honest advice is neither "avoid retail media AI" nor "embrace it uncritically." It's to insist on understanding what you're buying.

Ask your retail media partners what signals feed their AI targeting systems, what the attribution model is and how it handles multi-touch paths, and whether you can access the underlying audience logic rather than just the campaign outputs.

Some of those questions will get meaningful answers. Some will get polished non-answers. How a partner responds tells you something useful about how much they actually want you to understand.

The efficiency gains from AI-powered retail media are real. So is the risk that "the AI handled it" becomes a substitute for understanding what's actually happening to your media budget.