Let me start with something that shocked me: In December last year, Visa issued a statement with wording like this: "The 2025 shopping season marks the end of an era." A payment company suddenly started writing epic epitaphs. Visa announced that they had completed hundreds of real transactions initiated autonomously by AI agents, spanning their global partner network. Their SVP directly stated: "In 2026, AI agents won't just help you pick items—they'll complete the purchase for you."

My initial reaction to this news was: Another marketing piece, right? But after three months of careful research, I found that the scale of this is far beyond my expectations. This article is the result of my three-month research—What is Agentic Commerce? What's happening? What should overseas brands prepare for now?

What Exactly is Agentic Commerce?

Let me explain it in the simplest terms first. In the past, to buy a pair of running shoes online, you would: 1. Open a browser and search for "running shoe recommendations," 2. Scroll through dozens of reviews, 3. Open five websites to compare prices, 4. Checkout, fill in your address, and pay, 5. Wait for delivery. Throughout this process, you are the driver, and the platform is just a tool.

After Agentic Commerce, this scenario becomes: You speak to your phone: "Help me find a pair of trail running shoes, under $150, arriving by Friday, preferably not all black." And then... that's it. The AI agent automatically understands your intent, compares products across platforms, evaluates price, inventory, and shipping times, places the order for you, pays, and sends a confirmation email. Throughout the entire process, you never clicked "Add to Cart." You set the parameters; the Agent executes everything.

This is the essence of Agentic Commerce—AI autonomous agents complete the entire transaction loop from discovery to purchase. McKinsey predicts this model could redistribute $3 to $5 trillion in global retail consumption by 2030. Gartner believes AI agents will enable $15 trillion in B2B procurement by 2028.

Okay, the numbers sound sexy. But is it really happening now? Yes. And faster than you think.

What Happened in This Space in 2025-2026

Let me walk you through the timeline to feel the frantic pace of this industry.

2025: The Year of Infrastructure

September: OpenAI, together with Stripe, launched the Agentic Commerce Protocol (ACP) , an open AI commerce protocol that allows AI agents to interact with merchant systems and complete purchases in a standardized way. At the same time, OpenAI introduced the "Instant Checkout" feature in ChatGPT—US users could directly purchase Etsy items in the chat interface, with Shopify brands like Glossier, SKIMS, and Spanx following suit. Walmart opened up about 200,000 products for purchase within ChatGPT. At that time, ChatGPT had over 700 million weekly active users.

October: Visa, along with more than 10 partners including Cloudflare, Shopify, Stripe, Adyen, and Microsoft, launched the Trusted Agent Protocol—specifically addressing the identity verification issue of "is this order from a legitimate AI agent or a malicious bot?" The same month, Mastercard's CEO announced: "The first agentic transaction on the Mastercard network has been completed."

December: Visa announced the completion of hundreds of real end-to-end transactions initiated by AI agents. During the Christmas shopping season, e-commerce traffic from AI chatbots and browsers globally and in the US doubled compared to 2024, with AI credited for driving 20% of retail sales, contributing approximately $262 billion in revenue.

January 2026: Industry Acceleration

At NRF, the most important annual retail summit, the main theme this year was one word: Agentic. Google announced a joint launch with Shopify of the Universal Commerce Protocol (UCP) , a new open standard allowing AI agents to interact with any merchant's catalog and checkout process across platforms and systems using a unified language. Microsoft launched Copilot Checkout the same month, integrating with Shopify, PayPal, and Etsy. According to Microsoft data: after integrating Copilot Checkout, the rate of users completing purchases within 30 minutes increased by 53%, and shopping journeys shortened by 33% .

Shopify announced Agentic Storefronts—making products from its millions of merchants automatically available in conversational shopping scenarios across the four major AI platforms: ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini. No additional integration required for merchants, no separate app needed, zero development cost to start selling.

Do you realize what this means? Shopify unilaterally moved the shelves of millions of merchants worldwide into the AI chat interfaces that billions of people use daily.

March 2026: First Failure and Iteration

The industry took a breather. OpenAI quietly discontinued Instant Checkout. Why? The number of integrated merchants was extremely low (reportedly only about 30 Shopify merchants at launch), product data relied on web scraping leading to frequent price and inventory errors, merchant onboarding was complex, and the experience wasn't smooth. But this is not a failure of Agentic Commerce; it's the normal death of a first-generation product. Forrester analyst Emily Pfeiffer put it bluntly: "This isn't the death of agentic commerce; it's just that the experience isn't there yet."

OpenAI then pivoted: shifting its commerce focus to product discovery and dedicated apps (e.g., Target App, Instacart App, DoorDash App) . Users find products in ChatGPT and complete checkout in the brand's own app—merchants regain transaction control, while OpenAI remains the traffic gateway. Concurrently, Shopify announced an upgrade to Agentic Storefronts, allowing Shopify merchants' products to be discovered within ChatGPT and purchased via an embedded browser—shopping doesn't leave ChatGPT, but checkout happens in the merchant's own environment. This signals a more mature and sustainable division of labor is taking shape.

Understanding the Three Layers of Agentic Commerce

At this point, I find many people's understanding of this concept is muddled. Let me clarify with a framework. Agentic Commerce essentially has three layers:

Layer 1: Discovery

AI understands user intent and filters the most suitable options from a vast array of products. In this layer, ChatGPT, Google AI Mode, Perplexity, and Microsoft Copilot are all competing. The core question is: whose shelf is more complete, and whose understanding is more accurate. Adobe data shows that in mid-2025, AI-driven retail website traffic grew 4700% year-over-year. Shopify reported that orders from AI search grew 15 times throughout 2025. Consumers are already using AI to find things.

Layer 2: Decision

AI represents the user, comparing prices, shipping times, return policies, and product ratings across platforms, making recommendations or even direct decisions. This is the most "intelligent" part of Agentic Commerce: the agent isn't just searching; it's making judgments on your behalf . This means your products need to be readable by machines, not just seen by humans.

Layer 3: Execution

The Agent completes the actual purchase: placing the order, paying, and confirming logistics. In this layer, Visa, Mastercard, and Stripe are competing to build identity verification and payment infrastructure, while OpenAI's ACP and Google's UCP compete to set communication protocol standards. This layer is the hardest nut to crack. I'll explain why below.

Why is the Execution Layer So Hard?

The failure of Instant Checkout taught the entire industry a valuable lesson. The problem isn't that AI isn't smart enough; it's that the offline infrastructure isn't "agentic-ready." Most retailers' systems are designed for humans. Websites have various pop-ups, CAPTCHAs, and login walls that agents can't reliably penetrate. Inventory and pricing data rely on manual updates, so agents often scrape outdated data. Payment requires manual confirmation and identity verification, and agents lack legitimate identity authentication mechanisms.

The deeper issue is: When an AI agent places an order on behalf of a human, who bears responsibility? "Agents need an identity," Mastercard said. "You need to give it a secure identity, verify it, to ensure the entire transaction is trustworthy." This is the core technical challenge of Agentic Commerce today: establishing a trusted identity system for agents . That's why Visa's Trusted Agent Protocol, Google/Shopify's UCP, and OpenAI's ACP are all essentially solving the same problem—enabling systems to distinguish legitimate AI agents from malicious crawlers .

The Real Disruption: The Shift of Consumer Decision-Making Power

Now, I want to discuss a deeper issue than "who wins the protocol standard war." The most fundamental disruption of Agentic Commerce isn't the change in the shopping process; it's the transfer of consumer decision-making power from humans to machines.

Previously, your consumer decision chain was: Ad reaches you → You become interested → You actively search → You compare and choose → You order. Brands competed for your attention and mindshare.

After Agentic Commerce, this chain becomes: You set parameters → Agent searches for you → Agent compares for you → Agent orders for you . Brands now compete for the Agent's recommendation slot . This means your target customers are no longer just people—your target customers become the AI that helps people make decisions . The World Economic Forum describes this trend as: AI becomes the gatekeeper for brands. If brands don't proactively define their image in the eyes of AI, algorithms will define it for them. And the algorithm's definition may be completely different from the brand image you spent tens of millions building.

Three Things Overseas Brands Need to Understand Now

Alright, after all this research, let's talk about what's truly useful. As an overseas brand, with Agentic Commerce arriving, what should you do?

First: SEO is No Longer Enough; You Need AEO

SEO, Search Engine Optimization, has been the bible of internet business for the past two decades. But now, a new term has emerged: AEO—Answer Engine Optimization . What's the fundamental difference? The goal of SEO is to make your content rank higher , so more people click through. The goal of AEO is to make your content become the AI's answer , so AI directly cites and recommends you when responding to users. Traditional search engines give you a list of links; users choose. AI assistants give a direct answer and complete the purchase. When AI excludes your brand from recommendations, your traffic can drop to zero without users even noticing.

US Chamber of Commerce data shows that during the 2025 Christmas season, due to AI Overviews, organic click-through rates for informational search terms dropped 61% . Think about it: 61% of traffic, gone, and users didn't even realize they should have clicked your link. The core logic of AEO is: make your product data clear, trustworthy, and actionable for machines.

How to do it specifically?

  • Structured product data: Add Schema.org markup in JSON-LD format to product pages, including fields like GTIN, brand, price, inventory, shipping time, and return policy. This is the machine-readable language AI agents use to evaluate and filter you.
  • Natural language product descriptions: Your product copy should answer consumers' actual questions, not just stuff keywords. When someone asks ChatGPT "What lightweight running shoes for marathons under $200 do you recommend?", your product description should match that query.
  • Real-time accurate data: Price, inventory, and shipping times must be real-time. AI agents can't tolerate outdated data—outdated data will make agents drop you immediately.

Second: Get on AI Channel Shelves

The most immediate action is: put your products on shelves that AI can see. There are three main paths:

Path 1: Via Shopify's Agentic Storefronts. If you're a Shopify merchant, your products are already in the purchasable pool of ChatGPT, Copilot, Google AI Mode, and Gemini by default. The key is maintaining complete and real-time product data. If you don't have a Shopify store but want to enter this AI shopping ecosystem, Shopify has also launched a new service called Agentic Plan —even if you're not a Shopify platform merchant, you can sync products to the Shopify Catalog and leverage Shopify's AI channel infrastructure to reach consumers.

Path 2: Integrate OpenAI's ACP. OpenAI has open-sourced the Agentic Commerce Protocol, so merchants and developers can integrate directly. If your platform is Stripe-supported, you can enable agentic payment capabilities with as little as one line of code.

Path 3: Google's UCP. Target, Sephora, Nordstrom, Lowe's, and Best Buy have already integrated Google's Universal Commerce Protocol for product discovery. This channel is especially noteworthy because Google AI Mode's daily active users surpassed 75 million in March 2026.

Third: Rethink How Your Brand is Understood by AI

This is the hardest but most long-term task. When AI makes decisions for users, AI's understanding of your brand determines whether you get recommended. AI's understanding of your brand comes from three sources: 1. Your product data (website, product pages, product feeds), 2. Third-party content about you (media coverage, reviews, forum discussions, review platforms), 3. Your existing image in AI training data. You can't control #3, but you can actively manage #1 and #2. That is, design your AI Identity:

  • Your brand's information across all channels should be consistent and machine-readable.
  • You need to proactively build brand presence on high-authority media and citation sources (this affects how often AI cites you).
  • Your product positioning descriptions should be expressed in natural language dimensions that AI can understand, not just marketing jargon.

One metric worth watching: Citation Velocity —the speed and frequency with which your content is cited by major AI platforms. This is the most important brand exposure metric for 2026.

Several Underestimated Signals

Finally, let me share a few signals I think are worth attention but haven't been fully discussed.

Signal 1: Amazon is Fighting Back Fiercely, But May Not Hold

Amazon has been the biggest disruptor in Agentic Commerce so far—but not because it does it best, but because it's actively defending . Amazon has blocked dozens of AI agents from accessing its platform, including OpenAI's ChatGPT. Amazon even sued Perplexity in November last year to prevent its Comet browser from shopping on Amazon on behalf of users. Amazon's logic is clear: if consumers shop through AI agents, Amazon loses its direct relationship with consumers and the value of its advertising and recommendation systems. But how long can this defense last? An interesting detail: In February, Amazon invested up to $50 billion in OpenAI, and the two reached a strategic partnership. Bank of America analysts believe this may mean Amazon will eventually enter ChatGPT's commerce ecosystem in some form. The relationship between Amazon and Agentic Commerce will be one of the most compelling storylines to watch over the next 12 months.

(Image source: The AI Commerce Brief)

Signal 2: B2B Scenarios Will Land Faster Than B2C

We've discussed many consumer shopping scenarios, but one number stands out: Gartner predicts AI agents will enable $15 trillion in B2B procurement by 2028. This isn't an exaggeration. Think about corporate procurement: supplier price comparison, compliance review, order flow, payment approval—these processes are naturally suited for agents because they have clear rules, lots of repetitive actions, and high standardization. Visa has already started a real case in the Middle East: helping users automatically pay recurring fees like real estate service fees with AI agents. B2B overseas brands and platforms may feel the impact of Agentic Commerce earlier than B2C.

Signal 3: Logistics and Fulfillment Capability Will Become Key Factors for AI Choosing You

This is a completely underestimated signal. When an AI agent chooses a merchant for a user, what factors does it consider? Not just price and ratings, but also: Can it be delivered by Friday? When an AI agent evaluates two merchants selling the same product at the same price, it will choose the one that can provide real-time, accurate, machine-readable logistics information . A merchant that says "ships in 3-5 business days" is equivalent to non-existent in the agent's eyes. What does this mean for many overseas brands? Your logistics data transparency directly affects your ranking in AI recommendation results.

Final Thoughts: The Best of Times, and the Time for Early Movers

Christmas 2025 was the last holiday season where humans completed the entire shopping journey alone. This sounds philosophical, but behind it are real data and real business transformation. BCG's report predicts AI-led shopping will account for more than a quarter of total e-commerce spending. Adobe data shows AI-referred shoppers convert 31% better than other channels, with revenue per visit up 254% year-over-year.

But more importantly: It's still early in this race. Agentic Commerce is still in its first turn—protocol standards are still competing, product experiences are still iterating, and most brands haven't even figured out what AEO is. McKinsey's survey shows 62% of companies are experimenting with AI agents, but only 23% have begun scaling deployment. The window for early movers is often shorter than you think.

Google's e-commerce VP once said something I think every overseas brand should ponder: "If your brand information can't be read by AI channels, you won't appear in the AI recommendations that are driving sales." The next battlefield for going overseas isn't TikTok, not independent sites, not Google Ads. It's the Agent's recommendation list. Are you ready?