Your Retail Media Budget Just Got a New Line Item (and Your Dashboard Can't See It)

The deal that should be on every brand manager's radar

Earlier this month, The Grocer reported that Tesco-owned Dunnhumby has invested in Azoma, a UK startup that measures and improves how AI shopping agents discover, interpret and recommend products for retailers and brands. Azoma already works with L'Oréal, Unilever and Mars, and Arla and Mars were piloting the platform last year to get picked up by answer engines like ChatGPT, Perplexity and Google AI Overviews, plus retailer chatbots like Walmart's Sparky.

Here's why I think it matters. When the business that sits behind Tesco's loyalty data starts backing a tool for AI visibility, that's not a side bet. That's the retail data world telling you where the next fight is.

The stat that turns the playbook upside down

This is the bit that genuinely stopped me. Azoma analysed tens of millions of AI responses in Q2 2026 and found that earned or social media makes up 86.5% of the citations behind Alexa for Shopping's recommendations, and 76% behind Walmart Sparky's. ChatGPT leans more on retailer sources, at 37.1%.

Read that again. Your brand website, your lovingly signed-off product copy, your owned channels — they're a minority voice in the conversation that decides whether you get recommended. The agent is listening to what other people say about you: reviews, creators, press, Reddit threads.

Why your retail media dashboard is blind to this

Retail media reporting was built to answer one question: did the ad I paid for inside this retailer's walls get seen and clicked? It was never designed to tell you whether an AI assistant, pulling from sources all over the web, put your product on the shortlist.

Even dunnhumby's own ventures lead admits it — he said retailers and CPGs currently have no reliable way to link what an agent recommends to an actual sale. So we've got a channel that's already influencing baskets, with no standard measurement and no line on the media plan. That won't last. The minute measurement exists, paid placements follow — and the brands who've already built their earned footprint will be buying from a position of strength.

What I'd be doing now, not next year

You don't need an enterprise platform to get started. You need to stop treating this as someone else's problem.

•       Run your own audit — ask ChatGPT, Gemini and Alexa the questions your shoppers actually ask ('best high-protein yoghurt for kids', 'dairy-free snack for lunchboxes'). Screenshot where you appear, and where competitors do instead.

•       Invest in the earned layer — if 86.5% of citations come from earned and social, creators, reviews and PR stop being 'brand awareness' and become shelf presence. This is exactly where ongoing creator relationships earn their keep.

•       Clean up your product data — agents need to understand what you are. Consistent names, claims, ingredients and use cases across your site, retailer listings and feeds make you easier to recommend.

•       Extend your first-party data thinking — you already know which occasions and claims drive your shoppers. Make sure that same language shows up in the places agents read.

•       Ringfence a small test budget — not for AI ads yet, but for the content and review activity that feeds them, so you have a baseline before the paid formats arrive.

Owned channels aren't dead — they're just not enough

I'm not saying ditch your website or your retail media plan. Owned content is what the agent fact-checks against, and retail media still matters at the point of purchase. But the brands that win the next two years will be the ones who treat 'what everyone else says about us' as a channel they actively manage — not a happy accident. Channel integration is key - one channel isn't going to drive a sale, it's the journey of how the consumer gets there.