The Rise of Agentic Commerce: How AI Agents Are Reshaping Online Shopping

For decades, online shopping followed a predictable pattern. A consumer types a query into a search engine, clicks through a list of blue links, lands on a product page, evaluates options, and completes a checkout. That model is breaking apart.

In its place, a new paradigm is forming: agentic commerce, where autonomous AI agents browse, compare, negotiate, and purchase on behalf of consumers. These agents do not see banner ads. They do not respond to promotional emails. They parse structured data, evaluate trust signals, and execute transactions through machine-readable protocols. The implications for retail are enormous, and the transformation is already well underway.

The AI Agents Have Arrived

The shift from theoretical to operational happened fast. On January 23, 2025, OpenAI launched Operator, a web-browsing AI agent available to ChatGPT Pro subscribers at $200 per month. Powered by the Computer-Using Agent (CUA) model, Operator can navigate websites, fill forms, and complete purchases. Launch partners included DoorDash, eBay, Instacart, Priceline, StubHub, and Uber. By September 2025, OpenAI introduced Instant Checkout directly within ChatGPT, powered by the Agentic Commerce Protocol (ACP) developed with Stripe. Etsy sellers went live first; over one million Shopify merchants, including brands like Glossier, SKIMS, and Spanx, were next in the pipeline. With ChatGPT surpassing 800 million weekly active users, the commerce surface area is staggering.

Google moved with equal urgency. Project Mariner, unveiled at Google I/O in May 2025, demonstrated an agent scoring 83.5% on the WebVoyager benchmark while running ten parallel tasks for AI Ultra subscribers. Earlier, in January 2025, Google and Shopify co-developed the Universal Commerce Protocol (UCP), an open standard covering the full shopping journey from discovery to post-purchase support. Google's Agent2Agent (A2A) protocol, announced in April 2025 with over fifty partners, was contributed to the Linux Foundation by June.

Amazon took a different approach. Rather than building a standalone agent, the company embedded AI deeply into its existing marketplace through Rufus, its AI shopping assistant. With more than 250 million users and monthly active usage up 149% year over year, Rufus drives a 60% higher purchase likelihood among engaged shoppers. Analysts at Evercore project that Amazon's AI-driven advertising revenue could reach $4 billion by 2028, with an expected $700 million in direct profits from AI-assisted commerce in 2025 alone.

Microsoft entered the arena at the National Retail Federation conference in January 2026, announcing Copilot Checkout and Brand Agents for Shopify merchants, with payment processing through PayPal, Shopify, and Stripe. Perplexity launched Buy with Pro in late 2024, expanded to free shopping in November 2025, added PayPal integration, and introduced Snap to Shop visual search. Anthropic's Model Context Protocol (MCP), released in November 2024, became the de facto connectivity standard after OpenAI adopted it in March 2025.

This is not a single company's initiative. It is a coordinated, industry-wide structural shift.

Market Size and Growth Trajectory

The numbers tell the story of a market transitioning from nascent to inevitable. Grand View Research and Precedence Research estimate the AI agents market at $7.3 to $7.8 billion in 2025, growing at a 43% to 50% compound annual rate. Projections for 2033-2034 range from $139 billion to $183 billion.

Those figures capture the agent technology layer. The commerce impact is far larger. McKinsey projects that by 2030, AI agents will orchestrate $1 trillion in U.S. retail spending and $3 to $5 trillion globally. That projection reflects not just purchases made by agents, but the upstream influence AI exerts on product discovery, price comparison, and brand selection.

The traffic data confirms the trajectory is accelerating. Adobe Digital Insights reported a 4,700% year-over-year increase in AI-driven traffic to retail sites as of July 2025. AI-referred visitors convert at rates 31% higher than those arriving through traditional channels. Revenue per visit from AI-directed shoppers climbed 254% year over year during the 2025 holiday season. According to eMarketer, 26% of U.S. adults used AI for product discovery in 2025. McKinsey found that half of consumers now use AI during their search process.

For smaller businesses, adoption is climbing as well: 60% of small businesses reported using AI tools in 2025, an 18% increase from the prior year.

The Protocol Layer: Infrastructure for Agent Commerce

AI agents need structured ways to communicate with merchants. Four protocols have emerged to fill this role, each addressing a different layer of the commerce stack.

MCP (Model Context Protocol), released by Anthropic in November 2024, provides an open standard for connecting AI assistants to external data sources and tools. Its rapid adoption, by OpenAI, Google, Microsoft, and Shopify, has made it the foundational connectivity layer. When Shopify activated a default MCP endpoint on every store in summer 2025, it signaled that agent-readiness was no longer optional for the platform's merchants.

ACP (Agentic Commerce Protocol), announced by OpenAI and Stripe in September 2025, tackles the payment problem. Its key innovation is Shared Payment Tokens (SPTs), which allow AI agents to initiate purchases without exposing consumer payment credentials. These tokens are scoped to specific sellers and bounded by time and amount.

UCP (Universal Commerce Protocol), co-developed by Google and Shopify and announced in January 2025, covers the complete shopping journey: discovery, buying, and post-purchase interactions. With endorsements from over twenty major partners including Target, Walmart, Best Buy, Visa, and Mastercard, UCP aims to standardize how agents interact with commerce surfaces.

A2A (Agent2Agent Protocol), announced by Google in April 2025, enables secure collaboration between autonomous agents across different frameworks. By July 2025, version 0.3 added gRPC support and the consortium had grown to over 150 organizations.

These protocols are complementary, not competitive. MCP connects agents to data. ACP handles payments. UCP covers the shopping lifecycle. A2A lets agents coordinate with each other. But for merchants, each protocol represents a separate integration, and that is where complexity begins.

How Major Platforms Are Adapting

E-commerce platforms are racing to make their merchants agent-ready.

Shopify has been the most aggressive. Beyond co-developing UCP with Google, Shopify activated default MCP endpoints on every store during summer 2025 and launched the Shopify Catalog to give developers programmatic access to merchant product data. For Shopify's merchant base, baseline agent-readiness became a platform feature rather than a custom build.

WooCommerce published its AI and Agentic Commerce roadmap in October 2025, outlining MCP implementation and Stripe integration for its four million-plus stores. The open-source nature of WooCommerce means implementation varies significantly across merchants, but the direction is set.

On the payments and infrastructure side, Stripe launched its Agentic Commerce Suite in September 2025, introducing Shared Payment Tokens as a core primitive. PayPal followed in October 2025 with Agentic Commerce Services, including Agent Ready and Store Sync, tools designed to make PayPal-connected merchants discoverable to AI agents.

These platform-level moves are reducing the integration burden for merchants who operate within their ecosystems. Merchants on custom platforms, however, face a more complex path.

The Discovery Paradigm Shift: From SEO to AEO

Perhaps the most profound implication of agentic commerce is the transformation of product discovery. For twenty years, merchants optimized for search engine algorithms: keywords, backlinks, page speed, meta descriptions. That discipline, search engine optimization, shaped how billions of dollars in marketing budgets were allocated.

AI agents do not use search engines the way humans do. They consume structured data directly. They evaluate product attributes, compare prices programmatically, and assess trust signals like return policies, shipping reliability, and review aggregates. The new discipline is Answer Engine Optimization (AEO): making your store's data complete, accurate, structured, and machine-readable.

The performance difference is measurable. Research indicates that GPT-4's accuracy in understanding product information jumps from 16% to 54% when structured data is present. Merchants without schema.org markup, complete product identifiers (GTINs and UPCs), and machine-readable policies are effectively invisible to AI agents.

This is not a gradual evolution. It is a binary shift. Either an AI agent can read your store and recommend your products, or it cannot.

What This Means for Merchants

The strategic implications vary by merchant size and platform, but several themes are universal.

Structured data is now mandatory infrastructure. Product markup using schema.org vocabulary, machine-readable shipping and return policies, and complete product identifiers are baseline requirements for AI visibility. Merchants who treated structured data as an SEO enhancement must now treat it as core commerce infrastructure.

Protocol fragmentation creates real integration complexity. ChatGPT uses ACP. Google surfaces use UCP. Claude and a growing list of AI tools use MCP. No single protocol covers all agent surfaces. Merchants who want full AI visibility need to support multiple protocols, but building and maintaining separate integrations for each is resource-intensive, particularly for independent and DTC brands.

Trust verification is becoming a transaction prerequisite. AI agents are increasingly evaluating trust signals, business longevity, verified reviews, return rate data, operational metrics, before recommending merchants to consumers. This is a fundamental change from the SEO era, where ranking depended primarily on content and links.

AI traffic measurement requires new analytics approaches. Traditional web analytics tools were not designed to identify and attribute AI-referred traffic. Merchants need to distinguish between human visitors and agent-driven visits to understand how AI is influencing their sales.

Addressing the Complexity Gap

The gap between enterprise players and independent merchants is widening. Large retailers have the engineering teams to implement multiple protocols, optimize structured data at scale, and build custom agent integrations. A Shopify merchant with fifty products and no developer on staff faces a structurally different challenge.

This is the problem ORBEXA was built to address. As an AI-native infrastructure platform, ORBEXA provides multi-protocol support, covering MCP, ACP, UCP, and A2A, through a single integration. Rather than requiring merchants to build and maintain separate implementations for each protocol, ORBEXA exposes their product data, policies, and trust signals across all major agent surfaces simultaneously.

The platform also generates and maintains structured data in the formats AI agents expect, handles multi-language and multi-currency requirements for cross-border merchants, and surfaces trust verification data through the ORBEXA Trust Registry (OTR).

This is not about replacing the protocols. MCP, ACP, UCP, and A2A are industry standards built by Anthropic, OpenAI, Stripe, Google, and Shopify. ORBEXA implements all of them so that merchants do not have to implement each one individually.

What Comes Next

Agentic commerce is moving from early adoption to mainstream infrastructure. Several developments will define the next twelve to eighteen months.

Agent-to-agent transactions will grow in complexity. Today, most AI commerce involves a single agent acting on behalf of a consumer. The A2A protocol enables scenarios where a buyer's agent negotiates with a seller's agent, coordinating logistics, payment, and delivery without human intervention on either side.

Payment rails will mature. Shared Payment Tokens and similar constructs are still in early deployment. As trust frameworks develop and regulatory clarity improves, the volume of agent-initiated transactions will increase substantially.

The discovery layer will consolidate. Today, AI agents access merchant data through multiple protocols with overlapping capabilities. Over time, these will either converge or settle into clearly defined roles, reducing fragmentation for merchants.

Platform economics will shift. If AI agents, rather than human shoppers, drive an increasing share of purchasing decisions, the value of traditional advertising declines while the value of structured data, trust verification, and protocol compliance increases. Marketing budgets will follow.

The merchants who act now, investing in structured data, multi-protocol readiness, and trust verification, will be positioned to capture the next wave of commerce. Those who wait for the market to stabilize may find that by the time they are ready, the agents have already decided where to shop.