1 Google Marketing Live opens — the “Gemini advantage” becomes shipped product
Vidhya Srinivasan (VP/GM, Ads & Commerce) and Philipp Schindler (SVP/Chief Business Officer) opened Google Marketing Live 2026 at 8:45 AM PT today (May 20) as a global livestream. The framing: turn “the Gemini advantage” into a competitive edge for advertisers, structured around AI-powered campaigns, agentic commerce, and a new era of performance on YouTube. GML lands one day after the I/O keynote moved Gemini into the Android OS itself — meaning today’s Ads & Commerce announcements assume the OS-level surface is already there.
So what: This is the keynote that resets the next 12 months of paid media planning. Yesterday Google reset Search; today it reset the ad stack on top of Search. The point of the doublet is that Google is no longer pitching individual features — it’s pitching a new substrate where AI is the campaign manager, the checkout flow, and the creative agency, all at once. The practical takeaway for any marketing org: every line item in your 2026 plan has a new default vendor. The only real question is which parts of the stack you keep building yourselves vs. let Google handle.
2 Universal Commerce Protocol (UCP) launches — Shopify co-developer, Walmart, Target, Ulta, Wayfair, Etsy live
Google formally launched the Universal Commerce Protocol (UCP) — an open standard that lets AI agents act on behalf of consumers and advertisers across the entire shopping journey, including native checkout inside AI Mode and the Gemini app without a redirect. Shopify is Google’s protocol co-developer, and the launch wave of live retailers includes Walmart, Target, Ulta, Wayfair, and Etsy. Google said it will announce the next wave of retailers and a broader availability timeline in the coming weeks.
So what: UCP is the MCP of transactions — it standardizes how an agent sees inventory, prices, offers, and checkout. If you sell anything online and you’re not on the spec by Q4, agents literally can’t shop you. The retailer list also tells the strategic story: Shopify protects the long tail, Walmart/Target/Ulta/Wayfair/Etsy protect the megastores. Amazon is not on the list — which is the actual headline. For brands evaluating where to invest engineering this year, UCP integration just became more important than another SEO project. Push your eng team to read the spec this week, not next quarter.
3 Agentic Payments Protocol (AP2) + Universal Cart — Google’s payments standard pairs with UCP
Alongside UCP, Google introduced the Agentic Payments Protocol (AP2) — a complementary open standard for how agents authorize and execute payments on behalf of users — and a Universal Cart that sits on top of both. Together, the three components let a Gemini agent discover a product, build a cross-retailer cart, and check out — all without leaving the AI surface. Universal Cart is the visible UI; UCP is the inventory/offer layer; AP2 is the money-moves layer.
So what: This is Google formalizing what most retailers had been deferring: agents need a payment rail that isn’t credit-card-on-file inside every store. AP2 is Google’s pitch that the agent itself can hold tokenized credentials and disburse payments — which kills the “create an account on every site” friction that’s still the #1 e-commerce drop-off. The catch: AP2 routes payments through Google’s stack, which makes Google a payments rail company by default. For Stripe, PayPal, Shop Pay, and Apple Pay, the announcement reads like an opening shot — and the right response over the next 6 months is to ship competing protocols that any agent can speak.
4 Business Agent goes live in Merchant Center — Lowe’s, Michael’s, Poshmark, Reebok in wave 1
Business Agent — Google’s new branded conversational agent that retailers can spin up inside Merchant Center — goes live with Lowe’s, Michael’s, Poshmark, Reebok and others as the first wave of activated brands. Eligible U.S. retailers can customize the agent in Merchant Center, and in the coming months, train it on their data, surface customer insights, offer related products, and enable agentic checkout inside the experience. The agent meets shoppers directly on Search, “like a virtual sales associate,” before they leave for the retailer’s own site.
So what: This is Google quietly inverting the customer-acquisition stack. Until today, a brand’s job was to drive traffic from Search to its own site. Starting today, the brand’s job is to win the conversation inside Google’s surface itself — because the shopper may never click through. Business Agent gives retailers the ability to run the branded experience on Google’s real estate, but the trade is obvious: Google sees every conversation, and the algorithm decides which agent appears for which query. For brand teams, this is the moment to build a “Google-side persona” — the voice, copy, product catalog, and FAQ structure that a Business Agent will use to represent you. Treat it the same way you treated your homepage in 2015.
5 Direct Offers + new Merchant Center attributes — conversational commerce gets its data layer
Google also rolled out Direct Offers in Google Ads — retailers configure offers (discounts initially, with bundles and free shipping coming) in campaign settings, and Google’s AI decides when an offer is relevant to display in AI Mode, Gemini, and Business Agent surfaces. The other half of the announcement: dozens of new Merchant Center data attributes — answers to common product questions, compatible accessories, substitutes, and other conversational-commerce signals that go beyond keyword feed data. Merchant Center for Agencies is also rolling out globally.
So what: This is the operational layer behind UCP. If UCP is the protocol, Direct Offers and new attributes are what you actually configure on day one. The new attribute set is the bigger long-term story: it means shopping feeds are no longer just title + price + image — they’re question-and-answer data, accessory graphs, and substitution maps that AI agents use to reason about your product. For e-comm teams, the immediate to-do is to audit your product feed against the new attribute schema this week and update the highest-volume SKUs first. The brands that surface inside Gemini next month will be the ones who shipped feed data fastest.
6 Journey-aware bidding ships — long-cycle accounts finally win Search
Google announced journey-aware bidding for Search Ads as a GML headline product. The feature lets Smart Bidding learn from both biddable and non-biddable conversion goals simultaneously — meaning a Target CPA campaign can now use the full lead-to-sales funnel signal, not just last-touch conversions. Beta opened May 7, 2026. Google explicitly called out the winners: B2B SaaS, financial services, healthcare, higher education, and home services brands with sales cycles longer than two weeks.
So what: This is the bidding update lead-gen marketers have been begging for since Smart Bidding launched. Until now, a Target CPA campaign was effectively blind past the form-fill — which meant ad spend was optimized for cheap leads, not for revenue. With journey-aware bidding, the algorithm reads CRM-stage data and weights bids accordingly, so Google AI can finally tell a $200 MQL apart from a $20,000 closed deal. The practical move for lead-gen teams this week is to wire CRM-stage conversions into Google Ads via Enhanced Conversions or offline conversion imports. If you’ve been deferring that integration, the deferral just got expensive.
7 Demand-led pacing + Smart Bidding Exploration expand — to Performance Max and Shopping
Google paired journey-aware bidding with demand-led pacing — campaign budgets that automatically follow consumer demand inside a total monthly budget, spending more on peak days and less on slow ones. It builds on the campaign total budgets feature shipped earlier in 2026, which Google said reduced manual budget adjustments by 66%. Separately, Smart Bidding Exploration — the feature that lets Google temporarily relax your ROAS target to discover new queries — now expands beyond Search to Performance Max and Shopping campaigns. Google’s internal data: SBE drove an 18% increase in unique converting query categories and a 19% increase in overall conversions.
So what: Budget management just became a strategy decision, not an operational chore. Daily budget babysitting was always a bad use of media-team time, and most pacing software bolted on top of Google Ads was solving the same problem. With demand-led pacing, the budget itself becomes the constraint, not the daily cap — which means smaller teams can run more campaigns without losing nights to dayparting. SBE’s expansion to PMax and Shopping is the other story: the queries you weren’t winning are now part of the test set by default. The right calibration for 2026 budgets isn’t to set a tighter ROAS target — it’s to set a wider one, and let SBE find the queries your manual targeting would never have bid on.
8 Multimodal Video Creation — Gemini + Nano Banana + Veo write your video ads
Google introduced Multimodal Video Creation, an AI workflow that uses Gemini, Nano Banana, and Veo to go from written brief to final video ad in a few prompts. The pipeline is positioned as the production-cost killer for advertisers running video at scale across YouTube and Demand Gen — and it ships at a moment when YouTube reported 30% conversion lift on Demand Gen campaigns in 2H 2025 from creative enhancements. Early Demand Gen adopters are also seeing 3x higher CTR vs. paid social and 61% lower CPA.
So what: This is the moment creative production cost stops being the gating factor for video performance ads. A 10-person performance team can now ship 50 video variants a week without an external production house — and the math gets dangerous when you pair that with SBE: more variants × more queries = more conversions, at a marginal cost that approaches zero. The risk is that every video starts to look the same — Veo and Nano Banana have aesthetics, and at scale those aesthetics become a category visual. The strategic answer is to use Google’s video stack for the bottom 70% of creative volume, and reserve in-house and agency dollars for the top 30% that has to carry brand differentiation. Brands that get this split right will outrun the ones who go all-AI or all-human.
9 Walmart puts ads inside Sparky — retailer-owned agents become the parallel track
Walmart formally turned advertising on inside Sparky, its AI shopping agent — meaning sponsored placements now appear inside Walmart’s own conversational shopping experience. Walmart Connect is positioning Sparky as a retailer-owned agent distinct from Google’s Business Agent layer: shoppers stay inside Walmart’s app, see Walmart-owned conversation history, and pay through Walmart-owned rails. The move comes in the same week Walmart Connect’s Vizio integration crossed a full year of pairing first-party shopper data with connected-TV ad inventory.
So what: This is the retailer-owned agent counter-strategy to UCP, and it’s the more interesting story for any brand selling in big-box. Walmart isn’t joining Google’s protocol just to be one of many retailers — it’s running a parallel agent that owns the customer relationship and the data. Bain’s May 14 brief noted consumers trust retailer-owned agents 3x more than third-party agents to complete transactions, which gives Walmart real leverage. The implication for CPG and DTC brands: you’ll soon be running two parallel agentic playbooks — one optimized for Google’s UCP/Business Agent surface, one for Walmart-owned Sparky (and Amazon’s Rufus, and Target’s eventual answer). The brands that ship two clean product feeds first will eat the brands still running one.
10 Bain’s May 14 brief — agents disintermediate the top of the funnel
Bain & Company published a brief on May 14, 2026 arguing that agentic AI is restructuring the retail customer journey by disintermediating the top of the funnel — i.e., the discovery, comparison, and shortlisting stages are increasingly handled by third-party agents acting on behalf of the shopper, not by the retailer’s own marketing. Bain’s framing: brand-side marketing keeps the brand-equity job and the loyalty job; the agent eats search, comparison, and consideration.
So what: This is the strategic frame for every announcement above. If agents own the top of the funnel and retailer-owned agents own the bottom, brand marketers are squeezed into two narrow jobs: build enough brand equity that an agent volunteers your name when asked, and build enough first-party data that your retailer-owned agent can re-target the shopper after the agent hands them off. The 2026 budget question stops being “Meta vs. Google vs. TikTok” and becomes “Brand budget vs. agent-readiness budget.” Most marketing teams aren’t structured for this split — and the orgs that reorg first will have a 12-month head start on every competitor still running campaigns the old way.