Google’s 2026 developer conference wasn’t a feature list. It was a shift in the plumbing underneath how products get found, recommended and bought online. For ecommerce brands, a handful of those changes are worth acting on right now.
The Real Headline
Google shipped infrastructure, not features
Most coverage of a big launch event reads like a spec sheet. New model here, faster mode there, a few names you’ll forget by next week. That framing misses the point of what happened this year.
A feature is something you choose to switch on. Infrastructure is something that changes the ground everyone stands on, whether you opt in or not. Once a capability drops into the plumbing, every tool built on top of it inherits it by default. So does every competitor in your category. That’s the lens worth using here, because it changes what you do about it.
The brands that move from reading about this to wiring it into their store inside 90 days build a lead that gets harder to close every month.
Two Models, Two Jobs
One model is built to create. The other is built to act.
Google released two models that are easy to lump together and shouldn’t be. They solve two different problems inside a store, and confusing them leads to using the wrong one for the job.
|
Gemini Omni The one that makes things Turns any input into any output: video, image, text, audio. Feed it a product shot, get a campaign back. It doesn’t replace your creative lead. It deletes the three-week production timeline that used to sit between an idea and a finished asset. |
Gemini 3.5 Flash The one that does things This is the action model, not the chat model. Connect it to Shopify and it won’t hand you a draft to approve. It writes the listing and publishes it. That gap, between drafting and doing, is the whole story. |
The word carrying the weight for Flash is autonomous. You don’t walk it through every step. You set the goal, and it works out the path and executes it. For most teams, the mental shift is moving from “AI that helps me write” to “AI that finishes the task,” and those are not the same thing to manage.
The Agent Shift
Agents stopped being a demo and started being staff
For about two years now, “AI agents” has mostly meant slick demos that fall apart the moment they hit real conditions. Two announcements change that for teams that don’t have an engineer on call.
01 · Gemini Spark
An agent that acts on your behalf, around the clock
It watches inventory, flags low stock, drafts reorder messages and updates listings without being asked. I’ve been promising this kind of thing to clients for two years. This is the first version from a major platform I’d actually hand to a non-technical team.
02 · Antigravity
Build multi-step workflows with no developer
Google’s agent builder got a serious upgrade. You can chain steps together (pull the sales data, spot the trends, draft the weekly report) without touching code. This is the layer where a lot of the boring, repeatable work in a store quietly disappears.
Where Shopping Starts Now
A billion people are searching a different way
Google’s AI search mode has crossed a billion monthly users. Practically, that means a growing share of your customers now begin their shopping in a conversation, not on a page of blue links. Three things change because of that, and this is the section I’d act on first.
01 · The monitoring never sleeps
A shopper’s agent may already be watching your stock
When you restock a popular line, an agent can alert the customer before it even occurs to them to search. These agents already track apartment listings, sneaker restocks and product drops on their owners’ behalf. Your inventory status is now a signal someone else’s software is reading.
02 · Your product feed is your marketing now
Feed accuracy and schema stopped being a technicality
Merchant Centre data and structured schema used to live in the SEO weeds. They’re now front-line marketing infrastructure. If your feed has gaps, you’re simply invisible to the billion-plus people searching this way. Real-time accuracy and clean, structured product data are what these systems reward.
03 · Keyword density is finished
Intent and fresh reviews beat keyword stuffing
If your SEO approach still looks like 2018, you’re optimising for a game that’s over. What gets read now is structured product context and review recency, not how many times you crammed a phrase onto the page.
This isn’t theoretical anymore. Ulta Beauty has gone live inside Google’s AI shopping experience, letting shoppers get recommendations, compare and buy without leaving the conversation. When a retailer that size flips the switch, the question every board starts asking is: why aren’t we there yet?
Under The Hood
The boring layer that decides what’s possible
The infrastructure announcements feel more technical, and they matter precisely because they set the ceiling for everything else. Three worth keeping on your radar.
01 · Deep Research Max
Start here if you run an ecommerce team
Heavy data analysis pointed at your own business numbers. Ask it why average order value dropped last month and you get an actual answer, not a polite essay. This is the one that delivers insight rather than just output.
02 · Gemma 4
A capable open model you can run in-house
No API bill, and your data never leaves the building. That combination matters for any brand that wants full control over where customer information goes.
03 · Cheaper compute
The economics keep moving in your favour
New hardware means what costs a dollar today costs less next quarter. Every AI workflow you build gets cheaper to run over time, not more expensive. That’s a tailwind worth planning around.
What To Do With This
Pick one thing and wire it in
The mistake here is trying to adopt all of it at once. You can’t, and you don’t need to. Choose the highest-leverage move and connect it properly before reaching for the next one. For most stores, that move is your product feed and schema, because it’s the thing standing between you and a billion people who now shop through a conversation.
The brands that treat this as a reading exercise will still be reading in 90 days. The ones that pick a single change and put it into production will have built something the rest spend the next year trying to catch.
If you’re not sure which of these actually applies to your store, that’s worth a proper conversation. Book a call with Kelly and we’ll work out which one change is worth making first.
