Josh Uebergang.
Google is folding Shopping into AI Mode, and most Australian retail advertisers are budgeting as if nothing happened.
I have spent 20 years in ecommerce marketing and audited over 1,300 Shopify stores, and I can tell you the industry's centre of gravity is still media weight. More budget, more coverage, more bids. That logic is about to stop paying.
The unit of competition moves from the bid to the attribute
When a shopper asks an AI surface for "a waterproof hiking jacket under $300 that ships to Hobart this week", Google does not run an auction across everyone's keywords and hand the win to the biggest spender. It assembles an answer from product data. Price, availability, shipping speed, size, reviews, returns terms, all drawn from the merchant feed and the wider product graph.
The scale of that corpus is what most budget conversations are missing. Google told I/O this year that its Shopping Graph had passed 60 billion product listings, up from the 50 billion Sundar Pichai cited at the National Retail Federation in January, with more than two billion of them refreshed every hour.
Your catalogue is either legible to that or it is not.
If your feed says "Jacket - Navy - L" while a competitor's carries fabric, waterproof rating, fit notes and live regional shipping times, no amount of budget rebalances that fight. The model has nothing to say about you. Visibility now rests on the attribute values sitting in the feed, not on what a brand is willing to pay to be seen.
As others have argued here, you cannot optimise your way into an AI recommendation.
This is a quiet inversion of two decades of search economics. Since the early 2000s, money could paper over mediocre data because the ad unit was a blunt instrument. Ten blue links and a shelf of Shopping tiles rewarded whoever paid to be present. An AI answer rewards whoever is most useful to describe. Presence is no longer purchased, it is earned by the catalogue.
Who is exposed
Agencies whose retail search offer is bid management and budget stewardship are selling a skill the machine already absorbed. Smart Bidding took the bids years ago. AI shopping surfaces take the placement logic. What is left is work the industry has historically underpriced and understaffed, and which sits awkwardly inside businesses built around media. That is the uncomfortable version of the argument that AI is forcing agencies to remember what they are actually for.
Retailers with thin catalogues are exposed too. The brands that treated Merchant Center as a compliance chore, minimum attributes, disapprovals barely managed, GTINs missing, will find their products increasingly absent from the surfaces where buying decisions are being made. Absent is worse than outbid. Outbid at least shows up in an auction report. Absent is silent.
The reporting layer gets murkier
Click-based attribution already flatters brand traffic. AI answers compress the click path further, and the clicks are already falling. Pew Research Center found users clicked a traditional result about 8 percent of the time when an AI Overview was present, against 15 percent when it was not. SparkToro has the share of searches producing any click down almost ten percentage points across 2024 to 2026. The boards funding this spend will be staring at dashboards that explain less every quarter.
The advertisers who move to MER and contribution margin as their north star will keep making sane decisions. The ones still worshipping platform ROAS will be optimising a number that no longer describes reality.
So what actually shifts this year?
First, the budget conversation reweights. Every dollar debated for extra media should be debated against product data quality, because data now compounds and media does not. A dollar of enrichment works in every AI answer that follows. A dollar of media buys one impression.
Second, the commercial model has to catch up. Retail search is still largely sold as a percentage of media spend, which pays for budget growth and not for data quality. While scopes are priced that way, the incentive runs directly against the thing that now decides who wins the surface.
Third, product data stops being back-office hygiene. Feeds outsourced to a departed developer's script, attributes locked in a legacy PIM nobody can edit, shipping tables that do not reflect actual carrier reality, these are front-of-house marketing problems now.
Australia has less cushion
Our retail auctions are thinner, our category leaders are fewer, and a small number of well-fed catalogues can dominate an AI surface in a way that was harder when everyone got a tile on the shelf.
The Australia Post eCommerce Report put online spending at $82.6 billion in 2025, up 14 percent and now close to a quarter of all retail. The number underneath it is the more interesting one. The average online transaction has fallen to around $96, roughly $10 below where it sat in 2020, because shoppers are spreading the same money across more brands and more occasions. Discovery is doing more work than loyalty right now, which is the worst possible moment to be illegible to the surface doing the discovering.
The winners of the next three years of retail search are being decided now, in spreadsheets and product databases, not in bid strategies. The budget line that matters most in 2026 does not say media on it.
Josh Uebergang is the founder of Digital Darts, a Google Ads agency for Shopify stores, and author of Google Shopping for Shopify.