Blake Wright, Sales Director Brand & Agency at Shopfully, explains why the I shopping hype is outrunning the habit.
The numbers coming out of AI shopping look insane. Adobe tracked generative AI traffic to US retail sites up nearly 700 per cent year on year over the holidays, converting a third better than everything else, with AI touching a claimed fifth of global online holiday spend. Read enough coverage and you’d think the shopping list is already automated.
Then look at what’s actually happening here, and breathe.
IAB Australia’s new State of the Nation report shows where local AI investment really sits: back office. More than half of retail media networks are exploring AI, but for data analysis, targeting, measurement and reporting. What’s actually being bought is still display and search, and only 22 per cent of networks would call their own capabilities advanced. Our data says the same thing. Across 2.8 million monthly active Australian shoppers, the behaviour hasn’t moved: people open the app themselves, pick a store, build the week’s list.
Both things are true, and that’s not a contradiction. It’s a learning curve, and retail has run every one of them the same way. Newspapers trained generations to scan for what’s on and what’s cheap. The letterbox catalogue spent the better part of a century training the Thursday ritual. Even ecommerce, the fastest shift retail has ever seen, needed two decades to reach a minority of Australian sales. The tech shows up in a quarter. The habit takes years, then compounds. AI shopping will follow the habit curve, not the hype curve. So the right amount of panic is none. The right amount of preparation is more than most are doing.
There’s a second brake nobody at the AI conferences mentions: the loudest adoption numbers describe the smallest trolleys. Adoption gets counted in users, and users skew young. Spend gets counted in households, and the big weekly baskets belong to established families, whose habits are the stickiest in retail. The curve will be slowest exactly where the money is.
I used to sell rooms
Before retail media I sold magazines, and what we were really selling was never pages. It was rooms. An editor’s taste, built issue after issue, that readers chose to walk into. That choice was the targeting. It came with attention and trust already attached. Brands bought the room and got the whole audience, including people who weren’t buyers yet but were in the right frame of mind to become one.
That logic never died. It moved. An influencer is an editor with a phone. The follow button is a subscription. People still opt into someone’s taste, and they still trust what they find there.
But while audiences kept choosing rooms, advertising stopped buying them. Programmatic taught us to chase the individual across whatever environment they happened to be standing in, and once you’re chasing the person, the room stops mattering. So does the accidental discovery the room used to pay for. We got precision. We paid for it with everything we couldn’t measure.
And that sets up a question with a familiar shape, because I’ve written about the promo trap and the performance trap before. If precision targeting is the rational move for every single brand, who funds the discovery a category needs to grow? Chasing in-market audiences is each brand’s optimal play. Add it up across the market and nobody is creating new demand; the category quietly shrinks to the already-convinced. Price was the first race to the bottom. Performance was the second. Precision is the third, and it’s already running.
Fed is not the same as found
Now drop an algorithm, or an agent, on top of that already-narrowed system.
When a feed serves you a product, or an agent builds your basket, you haven’t discovered anything. You’ve been fed. The feed is built from your past. It’s retrieval dressed up as discovery. Actual discovery starts with the shopper: you open something to see what’s out there and you find things you weren’t looking for. The catalogue was a discovery engine precisely because everyone got the same forty pages and nobody had sorted them by relevance. The serendipity was the product.
This isn’t nostalgia, it’s arithmetic. Brands grow through light buyers and new buyers, the people whose baskets have never seen you. A feed trained on past behaviour rehearses what you already buy. So here’s the question the agentic pitch never answers: how does a challenger brand get into a basket built by an agent that has never seen anyone buy it?
Until someone solves that, the places where people still browse on their own terms get more valuable as the feeds narrow, not less. Out-of-home on the drive. Audio in the car. Total TV in the lounge room. The influencer a follower chose. And the pre-shop planning session, the Thursday catalogue’s direct descendant, where a shopper goes looking at what’s new and what’s on offer before they’ve decided what they want. These are rooms. People walk into them on purpose. A brand can still be found there by someone who wasn’t looking for it.
Where we stand, honestly
We’re not bystanders here. Shopfully moved past putting the catalogue on a screen years ago. We surface the right offers, brands and products to the right shopper while they’re planning, and yes, there’s machine intelligence in that. The difference we’d defend is who starts the session and why: the shopper opens it, it’s about the week ahead, and the surface is built to introduce as well as retrieve. Do we always get that balance right? Fair question. We’d rather be asked it than duck it.
AI’s place is real. When agents do start building baskets at scale, they’ll need exactly what this channel has spent years structuring: current, local, machine-readable offer data. But that’s plumbing for a behaviour that’s still forming, in the households that haven’t formed it. The call for brands right now is older and simpler. While the new habit trains itself, keep buying rooms. Because once everything is fed, being findable is the only advantage left.
People don’t discover what they’re fed. They discover what they go looking for. And they still go looking in rooms someone cared enough to curate.
Sources: Adobe Analytics holiday 2025 US retail traffic data; IAB Australia Retail and Commerce Media State of the Nation 2026; Mi3 IAB summit coverage, July 2026; ShopFully first-party shopper data, 2025-26. To verify at submission: ABS household expenditure by age cohort for the basket-weight claim; Adobe figures against the primary release.
Keep up to date with our stories on LinkedIn, Twitter, Facebook and Instagram.