Google Marketing Live 2026: Gemini Now Runs the Ad Stack, and Shopping Just Went Agentic

Pull up Merchant Center and open your five best sellers. Read each listing the way a machine would, cold, no benefit of the doubt. Could you recommend that product for a specific need based only on what is in the feed? If the answer is no, that is your real problem this quarter, and it just got more urgent.

Because Google changed the rules in May. At Google Marketing Live, the company stopped bolting AI onto its ad products and rebuilt the stack around Gemini. Strip the keynote theater and the pitch is this: an agent now sits between your catalog and the buyer, doing the discovery, the comparison, sometimes the checkout. For a seller that moves two things at once, where people find your products and how much creative you need to keep the machine fed.

Most of the recaps filed it under agentic commerce and moved on. Fine. The narrower question is the useful one. If a Gemini agent is running point on the shopping journey, what do you actually change this quarter? Four of the announcements move the needle. The rest was stage lighting.

The four that actually touch e-commerce

Shopping ads now live inside AI Mode

Start with the big one. AI-powered Shopping ads now show up inside AI Mode, where Gemini answers a shopper conversationally instead of handing back ten blue links. Someone asks for a lightweight running shoe for flat feet under 120, and the agent can surface products with a short per-product reason pulled straight from your feed. Your listing competes on how well its data answers the question, not just on bid. If you have been treating the feed as an afterthought, that is the wake-up.

Ask Advisor: one agent across the whole stack

Then Ask Advisor. Google folded its scattered in-product assistants into one Gemini agent that spans Google Ads, Analytics, Merchant Center, and the Marketing Platform, with a shared memory so it carries context across them. Ask which campaigns are leaking budget and it answers using Analytics and Merchant Center data in the same thread. Beta, English first. For a small team with no in-house PPC analyst, it is the most immediately useful thing they showed.

Asset Studio: Gemini Omni and one-click testing

Asset Studio got the Gemini treatment too: multimodal generation, video through Gemini Omni, and a one-click creative testing flow that spins variants and tells you which ones won. Google says it starts rolling out over the summer, English first. If you already build assets in Google's tooling, the cost of making more of them just dropped. We have a ground-level walkthrough of the older Asset Studio workflow if you want the version before the Gemini layer lands.

Universal Commerce Protocol, now in more places

The fourth is the plumbing. Google is pushing its Universal Commerce Protocol into Demand Gen, YouTube Shopping, and AI Mode. Plain version: the path from a product showing up to a buyer actually paying is getting shorter and more automated across more places. Fewer manual steps between discovery and checkout means the creative and the feed carry more of the persuasion, because there is less in between for a shopper to bail on.

What to change in your feed and creative now

You do not need to rebuild your account this week. You do need to treat the product feed as a ranking surface, because that is what it became on May 20. Start with completeness. An agent answering a specific question can only pull products whose attributes are actually filled in: material, fit, use case, dimensions, compatibility, the boring fields nobody enjoys. A listing with a title, a price, and nothing else is invisible to a Gemini answer trying to match intent. Audit the feed for blank or thin attributes and fill the ones that map to how buyers describe what they want.

Next, the product page itself: clear specs, real reviews, machine-readable markup. The agent reads the page the way a careful shopper would, then summarizes. Give it something accurate to summarize. Then the creative. Conversational surfaces reward a clean product shot and one tight value line over a busy banner, every time.

The creative-volume problem nobody put on a slide

Here is the part nobody put on a slide. More placements, more agentic surfaces, and one-click testing all point the same way. You need more creative variations than you used to, and you need them faster. One hero image and a single video does not fill Demand Gen, YouTube Shopping, AI Mode, and standard Shopping with the variety those systems chew through in a week.

Across Coinis accounts, the brands that keep up are not the ones with the biggest design teams. They are the ones that stopped making two ads a week by hand and started generating ten or more, then let the platform pick winners. The math is not subtle. If Google can test and place creative almost automatically, your bottleneck moves upstream to production, and whoever produces on-brand variations fastest gets the most shots on goal.

Feed the machine without hiring a design team

That is the gap Coinis was built for. Paste a product URL, our AI ad generator spins image, video, and UGC-style variations, and the e-commerce workflow launches them to Google, Meta, and TikTok from one place. When Google adds another agentic surface, you are not scrambling to brief a designer, you are pushing ten more variants live the same afternoon.

Gemini running the ad stack is not a threat to small sellers. It tilts the field your way, but only if you can feed it the volume of good creative and clean feed data it needs to pick you. Treat May 20 as a production problem, not a panic. Try Coinis AI and put your feed-to-launch on the same speed Google just handed its agents.