Agentic Commerce: What Happens When AI Does Your Shopping

A few weeks ago, in our issue on AI agents, we mentioned a term in The Jargon section ("agentic commerce") and promised we'd cover it soon. In the last few weeks, this topic has gotten bigger.

Amazon sued an AI startup for sending shopping bots to its website. OpenAI signed deals with Walmart, Target, and Etsy to let people buy things without ever leaving ChatGPT. Google launched its own AI-powered checkout. And during the 2025 holiday season, AI tools influenced $262 billion in global online sales—roughly 20% of all digital commerce over the holidays.

Something fundamental is shifting in how we buy things.

Right now, online shopping means you do all the work. You open a browser, search for a product, scroll through results, read reviews, compare prices across tabs, enter your credit card, click buy.

Agentic commerce flips that. You tell an AI "I need a winter jacket, my style, under $200, fast shipping" and it handles the rest—searches dozens of retailers, analyzes reviews, checks for sales, and either recommends the best option or just buys it for you. You never see a product page.

That's the shift: from you navigating the internet to an AI navigating it on your behalf.

The infrastructure went from concept to live commerce in about twelve months.

In September 2025, OpenAI launched Instant Checkout—a feature that lets people buy products without leaving a ChatGPT conversation. Etsy was first. Within weeks, Walmart, Target, and over a million Shopify merchants signed on. You can ask ChatGPT for a meal plan and buy all the ingredients from Walmart without opening a browser tab.

Amazon took a different—and revealing—approach. It launched "Buy for Me," a feature that uses AI to purchase products from other retailers' websites while you stay inside the Amazon app. Its AI assistant, Rufus, served over 300 million customers in 2025 and drove nearly $12 billion in sales.

But here's the twist: while Amazon was building its own AI shopping tools, it was simultaneously blocking everyone else's. It barred AI bots from OpenAI, Google, and Anthropic from accessing its site. Then in November, it sued Perplexity AI, accusing the startup's browser of secretly shopping Amazon on behalf of users.

Amazon wants to be the only AI that shops Amazon. Everyone else wants to be the AI that shops everywhere. Perplexity's response to the lawsuit said it all: a blog post titled "Bullying is not innovation." That tension defines the whole story.

And it's not just retailers and AI companies. PayPal, Stripe, Visa, and Mastercard all launched agentic commerce products in 2025—racing to build the payment rails that let AI agents transact securely. When the companies that move money are retooling their infrastructure for AI shoppers, you know the shift is real.

Every era of commerce has been defined by a single question: who controls how customers find products?

The Sears catalog controlled it in the early 1900s. Shopping malls controlled it in the 1970s. Google search controlled it in the 2000s. Amazon's search bar controlled it in the 2010s. In each case, whoever sat between the customer and the product captured enormous power—and profit.

AI agents are the next front door.

When you ask ChatGPT for running shoes, OpenAI decides which products you see—not Google, not Amazon, not Nike's marketing team. OpenAI says it ranks results by relevance, not by ads or paid placement. If that holds, it's a radical shift from the ad-driven model that has powered online retail for two decades.

Here's what makes agentic commerce truly different from previous shifts: AI agents don't see ads, don't feel brand loyalty, and don't impulse buy.

When you browse Amazon, you might go looking for laundry detergent and leave with a new set of earbuds because the deal caught your eye. Amazon's entire business model—$50 billion in annual advertising revenue—depends on you doing exactly that. An AI agent won't. It buys the detergent and nothing else.

But the disruption doesn't stop at advertising. Pull the thread and it unravels across the entire economy.

Brand strategy. Decades of emotional marketing—the Super Bowl ads, the Instagram campaigns, the carefully crafted brand stories—matter less when the buyer is a bot evaluating price, ratings, and product data. Companies are already shifting budgets from SEO to what they're calling AEO: Answer Engine Optimization. The goal isn't ranking high in Google anymore. It's making sure an AI agent can understand and recommend your product.

Pricing. When every AI agent comparison-shops instantly across every retailer, price transparency becomes total. Margins compress. The race to the bottom accelerates. Brands that relied on customer laziness or loyalty to maintain premium pricing lose that cushion.

Retail itself. If product discovery moves to AI, what happens to the storefronts, sales teams, and merchandising strategies designed to influence purchase decisions in person and online? The entire infrastructure of persuasion—from shelf placement to pop-up ads—starts to lose its purpose.

This is why the stakes feel existential to retailers. It's not one thing changing. It's the foundation shifting under everything at once.

Here's the uncomfortable question at the center of all this: how do you trust an AI with your money?

Amazon's "Buy for Me" hands your encrypted credit card to an AI that fills it into a third-party website. What if it buys 1,000 pairs of socks instead of 10? When an AI with your payment credentials is roaming the open web, every website becomes a potential vector to manipulate it. Nearly 80% of financial institution leaders expect fraud to increase as a result of agentic commerce.

The industry is scrambling to build guardrails. Visa and Mastercard are developing tokenized credentials—one-time-use payment codes so your real card number never touches a third-party site. Think of it like a hotel key card that expires after checkout instead of handing over your house key.

But money is only half the problem. The other half is identity. When an AI agent shows up to buy something, how does the merchant know it's actually authorized to act for you? There's no "agent passport" yet—no standard way to verify that this bot is legitimately shopping on behalf of a real, approved human. That's a massive unsolved gap.

The likely solution is a layered system: spending caps per transaction, human-in-the-loop approval above certain thresholds (your agent auto-buys your usual coffee order, but flags a $2,000 laptop for your confirmation), and restricted categories where the agent simply can't act alone. The companies building these guardrails—the payment processors, the AI platforms, the retailers—are effectively writing the rules for how billions of dollars will move. And they're doing it right now, mostly behind closed doors.

Despite the momentum, a reality check: ChatGPT drove less than 1% of e-commerce sessions over Thanksgiving weekend. About half of consumers say they're not comfortable letting AI make purchases on their behalf. McKinsey describes an "automation curve" from Level 0 (subscribe-and-save) to Level 4 (AI managing your household budget autonomously)—and most people are barely at Level 1.

But here's why it matters now: the infrastructure is being built today, and the companies building it will define the rules. By the time most people are comfortable letting AI buy their groceries, the power structures will already be locked in—just like they were with Google search and Amazon's marketplace.

The race isn't for your shopping cart. It's for the right to sit between you and every purchase you'll ever make.

If you work anywhere near marketing, sales, or product, you're going to hear this term a lot this year.

For two decades, businesses have obsessed over SEO—Search Engine Optimization—making sure their website ranks high when someone Googles something. AEO is the new version of that game, but for AI.

When an AI agent shops on your behalf, it doesn't see banner ads or branded landing pages. It reads product data—specs, reviews, pricing, availability. AEO is the practice of making sure your product is structured so an AI agent can find it, understand it, and recommend it.

Drop this in conversation: "The companies that win won't be the ones with the best Google ranking. They'll be the ones whose product data an AI can actually read. That's why everyone's shifting from SEO to AEO."

Next time someone mentions online shopping or AI, here's your move:

If someone asks why they should care:

McKinsey's "The Automation Curve in Agentic Commerce" (January 2026) maps out the five levels of AI shopping—from basic subscriptions to fully autonomous agents managing your household budget. It's the clearest framework for understanding where this is heading and how fast. Worth bookmarking for the next time someone asks "but how real is this?"

That's Gist for this week. See you next week.