Agentic commerce reshapes paid media as AI agents start buying

The old Google Ads playbook assumes a person types, compares, clicks, and buys. Agentic commerce breaks that chain, because AI agents are now doing the comparing, shortlisting, and sometimes the checkout on behalf of the shopper. That is not a theoretical shift, either: Salesforce says AI and agents influenced 20% of Cyber Week purchases in 2025 and drove $67 billion in sales.

The market has already crossed the line

The cleanest warning sign is the scale. Salesforce said its Cyber Week 2025 analysis covered November 25 to December 1, based on shopping data from more than 1.5 billion shoppers, and global sales reached $336.6 billion, up 7% year over year. It also said it powered 61 million orders on Agentforce Commerce during that week with 100% uptime. When that much money is flowing through agent-mediated paths, treating AI shopping as a side experiment is just bad media math.

That scale changes the job of the paid media team. You are no longer optimizing only for queries typed by a human with a search bar in front of them. You are optimizing for a system that can compare offers, price, shipping, returns, and product specs before a person ever sees an ad or product page. If your strategy is still built around keyword intent alone, you are missing the layer that now decides whether a product even gets shortlisted.

Product feeds are becoming the new bidding language

In agentic commerce, the feed is no longer just a catalog. It is the source of truth that teaches the agent whether your offer is worth surfacing. That makes structured data about price, availability, shipping speed, returns, and specifications far more important than the old habit of stuffing everything into generic ad copy and hoping the landing page picks up the slack.

Google’s Universal Commerce Protocol, or UCP, is a good sign of where the market is going. Google describes UCP as an open standard for turning AI interactions into instant sales, and says it is meant to enable agentic actions in AI Mode and Gemini, starting with direct buying. Google has also said it introduced a Universal Cart and expanded UCP across its shopping surfaces, which tells you the company sees agentic checkout as part of the core commerce stack, not a novelty tucked into a lab.

For agencies, that means feed management becomes bidding strategy. Price mismatches, stale inventory, broken shipping promises, or vague attribute data are no longer just catalog hygiene issues. They are reasons an agent may skip your product entirely.

Creative has to persuade both the buyer and the machine

Traditional paid search copy was written to trigger a click from a person with a vague intent signal. Agentic commerce forces a second audience into the room: the software deciding which products make the cut. That means the creative brief has to carry more precise claims, cleaner product language, and fewer fluffy superlatives that agents can’t verify.

The smartest message now is specific and machine-readable. If the offer ships in two days, say so. If the return window is generous, surface it. If there is a compatibility angle, put it front and center. Agents are built to compare inputs, not admire branding poetry, so the creative that wins will be the one that compresses the decision tree instead of decorating it.

Google’s 2026 Merchant Center updates point in the same direction. The company said Merchant Center now adds AI-powered insights and recommendations based on product and performance data, and it added AI performance insights and conversational attributes so retailers can better understand how products appear in AI shopping experiences. That is a clear signal that creative is being judged in an environment where product metadata and performance signals are being read together.

Landing pages and checkout paths need to be agent-ready

This is where a lot of brands will get exposed. If an AI agent can get all the way to a purchase decision and then hits a messy checkout flow, the whole optimization chain falls apart. The landing page can no longer be a glossy brochure that relies on human curiosity to carry the sale; it has to answer the questions the agent is already asking.

That means tighter product detail pages, fewer dead ends, and a checkout path that behaves consistently across surfaces. Google has already tied AI Mode, Gemini, Search, Maps, Merchant Center, and the Universal Cart together in its 2026 shopping and ads materials, saying the tools help retailers sell across AI surfaces. OpenAI and Stripe have pushed in the same direction with Instant Checkout in ChatGPT in the U.S., starting with Etsy merchants and moving to Shopify merchants soon, built on the Agentic Commerce Protocol they co-developed. OpenAI’s pitch is blunt: “ChatGPT doesn’t just help you find what to buy, it also helps you buy it.”

The implication is simple. If the buying journey can begin in one assistant, continue in another, and complete through a protocol-driven checkout flow, your old page-by-page funnel maps start looking fragile fast.

Measurement is about to get much messier

The hardest part for agencies will be attribution. Different checkout paths, different assistant behaviors, and different surfaces will all distort how conversions are credited. A click path that used to be obvious may now be split across an AI result, a cart handoff, a merchant feed, and a protocol-based checkout.

That is why conversion integrity matters more than ever. You need clean event tracking, consistent product identifiers, and a way to compare performance across surfaces that do not behave like classic search results. If you cannot tell whether a sale came from a human-led search, an AI assistant, or a hybrid path, you will overvalue some campaigns and underfund the ones actually driving business.

Google’s tooling suggests the company knows this will be a headache. Merchant Center is increasingly positioned not just as a feed manager, but as a measurement and insight layer for AI-driven commerce. That should be the cue for agencies to audit their own tracking stack now, before agent-driven traffic starts making the reports harder to trust.

The commerce stack is being rebuilt around agents

This shift is not coming from one company alone. Shopify has said UCP has broad support from Walmart, Target, Etsy, American Express, Mastercard, Stripe, and Visa, and it said some Meta experiences will soon use UCP for checkout. That kind of backing matters because it suggests the standards around agentic commerce are beginning to harden across merchants, payments, and platforms at the same time.

OpenAI, Stripe, Google, Shopify, and the major retail and payments players are all signaling the same thing: the commerce layer itself is being reorganized around agents. For agencies, the practical takeaway is harsh but useful. Tighten your feeds, rethink keyword strategy around shortlisted products instead of just search queries, write creative that survives machine scrutiny, simplify landing pages, and fix measurement before the new traffic patterns expose every weak link. The teams that adapt early will get the best advantage in a market where the buyer is no longer always human at the first touch.