Key Takeaways
- An open camp has formed around Google's UCP and OpenAI's ACP, with 20+ retail and payment companies participating
- Amazon is blocking AI agents to protect $68.6 billion in ad revenue, standing alone as the only major player pursuing a closed strategy
- The structural tension between "protocol standardization vs platform dominance" will shape merchant strategy for years to come
Open vs Walled Garden — What Is Happening in Agentic Commerce
In January 2026, an unusual scene unfolded on the NRF exhibition floor in Las Vegas. Google, Shopify, Stripe, Visa, Mastercard, and Salesforce — companies that compete fiercely with one another — rallied under the same protocol banner. Meanwhile, at the Amazon booth, representing roughly 40% of the US e-commerce market, there was no mention of open protocols whatsoever.
The world of agentic commerce is splitting into two distinct camps. One side envisions an "open" future where AI agents can freely traverse merchants. The other defends a "walled garden" that keeps everything inside a proprietary ecosystem. This clash is not merely about technical standards — it is a power struggle that will reshape the very structure of commerce.
Why Amazon Is Building Walls
The answer to this question can be distilled into a single number: $68.6 billion. That was Amazon's advertising revenue in 2025. Q4 alone generated $17.7 billion, up 24% year-over-year — growth that outpaced even AWS in propelling Amazon's bottom line.
In a world where AI agents shop on behalf of consumers, nobody "sees" sponsored listings. While Rufus sessions boost purchase completion rates by 3.5x, removing ads from that buying process would undermine Amazon's fastest-growing revenue stream.
That is why Amazon blocked ChatGPT and Perplexity crawlers via robots.txt in November 2025. In March 2026, Judge Maxine Chesney issued a preliminary injunction blocking Perplexity's Comet browser under the CFAA. Amazon built not just technical barriers but legal barricades as well.
Is Amazon ignoring agentic commerce, then? Far from it. Rufus has 250-300 million users, with monthly active users up 149% year-over-year. The "Buy for Me" feature provides access to over 100 million products from 400,000+ merchants — all within the Amazon app. Amazon is implementing agentic commerce inside its own castle walls.
The logic is straightforward. Letting external AI agents in would erase ad revenue. Growing proprietary agents lets Amazon evolve the customer experience while preserving the ad model. For Amazon, openness is synonymous with giving up profit — which is exactly why it has joined neither UCP nor ACP.
The Open Camp Assembles — Why Rivals Are Joining Forces
On January 11, 2026, Google CEO Sundar Pichai took the NRF stage to announce the Universal Commerce Protocol (UCP). Multiple media outlets reported the room shifted. Shopify, Walmart, Target, Best Buy, Macy's, Etsy, Wayfair, The Home Depot. On the payments side: Visa, Mastercard, American Express, Adyen, Stripe. More than 20 companies endorsed the open standard.
Why did so many companies rally under Google's banner? The driving force is a shared concern about Amazon's dominance.
Walmart's actions reveal the underlying calculus. The company deployed its internally developed AI agent "Sparky" across every external channel available — ChatGPT, Gemini, Google Search's AI Mode. When OpenAI's Instant Checkout stumbled due to accuracy issues, Walmart embedded Sparky directly on that channel, reportedly achieving purchase completion rates of roughly 70% compared to Walmart.com direct visits.
The OpenAI camp also raised its own flag. Agentic Commerce Protocol (ACP) was announced in September 2025 and published as open source from day one. While UCP covers the full purchase lifecycle from discovery to post-purchase support, ACP specializes in the checkout experience. It is tightly coupled with Stripe's payment rails, designed to complete purchases within ChatGPT's conversational interface.
Critically, UCP and ACP are not an exclusive format war like VHS vs Betamax. UCP focuses on discovery and scale; ACP focuses on conversation and intent understanding. Both are designed to coexist, and Shopify already supports both protocols. There is competition within the open camp, but the parties are aligned on one point: fighting outside Amazon's walls.
| Comparison | Open Camp (UCP / ACP) | Closed Camp (Amazon) |
|---|---|---|
| Product Discovery | Multiple AI agents search and compare across merchants | Rufus and Alexa+ operate within Amazon only |
| Payments | Multiple providers: Visa, Mastercard, Stripe, etc. | Amazon Pay-centric proprietary payments |
| External Agent Access | Standardized via APIs and protocols | Crawlers blocked via robots.txt |
| Ad Model | Exploring new ad formats via agents | Defending existing sponsored ads ($68.6B) |
| Partner Count | UCP alone has 20+ participants | Amazon standalone (40K+ merchants via Buy for Me) |
| Merchant Freedom | Multi-channel presence simultaneously | Subject to Amazon's rules and fees |
The Reality Merchants Face
Understanding the macro picture is one thing. The reality e-commerce operators face is far more messy.
In March 2026, Shopify opened Agentic Storefronts to millions of merchants. ChatGPT, Microsoft Copilot, Google Search AI Mode, and the Gemini app — all manageable from a single Shopify admin dashboard. The new Agentic plan even allows brands without a Shopify store to list on AI channels through Shopify's catalog infrastructure.
In stark contrast, among merchants incorporated into Amazon's Buy for Me, more than 180 businesses reported their products were listed without consent. Opting out requires emailing branddirect@amazon.com, with opt-in as the default.
Yet the choice is not so simple as "go open." Amazon's Rufus users are 60% more likely to complete purchases. Few businesses can afford to ignore a platform that accounts for roughly 40% of US e-commerce. In practice, merchants are forced into a two-front strategy: expanding external channels via open protocols while simultaneously optimizing for Amazon's internal algorithms.
According to Stripe's 2026 NRF report, approximately 75% of retail technology leaders are implementing or planning agentic commerce integration. Yet there is no consensus on which standard to adopt. Merchants bear the cost of dual integrations, agents must choose default rails, and innovation slows as resources split between protocols — this is the unvarnished reality of spring 2026.
The Ad Model Collision — The Wall Protecting $68.6 Billion and the Experiments Beyond It
The sharpest edge of this conflict lies in its impact on advertising.
Amazon's motivation for building walls is clear, but the open camp has not yet established a viable "commerce without ads" business model either. As Klaviyo's analysis points out, the structure where AI systems sit between brands and consumers is fundamentally at odds with traditional advertising models. Consumers ask AI agents before visiting brand sites, and "many of the most important decisions are already made" before any direct brand engagement.
What concerns brands most is the information asymmetry — "seeing outcomes without understanding why." Knowing what sold but losing insight into why, as those insights get absorbed into AI platform black boxes. This dynamic can breed new walled gardens even within the open camp.
Google is showing signs of embedding a commerce media suite within UCP, exploring AI-mediated ad formats originating from YouTube and Google Search. Criteo also announced an advertising partnership with ChatGPT. The open camp is not abolishing ads — it is building new forms of advertising mediated by AI agents through trial and error.
Which Side Will Prevail
History offers one useful analogy. In the PC era, open IBM-compatible machines overwhelmed Apple's closed model. But in the mobile era, Apple's iOS built an ecosystem rivaling Android, proving that a "closed world" could sustain a highly profitable business. In e-commerce, Amazon has similarly established a layered moat — Marketplace, Prime, FBA, and advertising — that makes external siege extremely difficult.
McKinsey predicts agentic commerce will reach $1 trillion in US retail by 2030. Morgan Stanley projects that roughly 50% of American shoppers will use AI agents by the same timeframe. As the market expands, the network effects of open protocols accelerate. The more merchants join the UCP ecosystem of 20+ participants, the more valuable it becomes for agents, attracting still more merchants in a virtuous cycle.
Yet inside Amazon's walls, there remain 250-300 million Rufus users and a catalog exceeding 100 million products. At this scale, Amazon can sustain its own agent economy without joining external protocols.
In the short term, Amazon's closed strategy will protect ad revenue and maintain margins. Over the medium to long term, however, as NIST launched its AI Agent Standards Initiative in February 2026, regulatory and standardization forces are likely to bolster the open camp. From a merchant's perspective, supporting open protocols is becoming not "insurance" but a "required investment."
What E-Commerce Operators Should Do Now
There is no luxury of waiting to see how this power struggle plays out. With 75% of retail technology leaders already in motion, inaction itself is the risk.
Prioritize structured product data — UCP, ACP, and Amazon's internal agents all assume machine-readable product information
Start with one open channel via Shopify Agentic Storefronts or UCP-compatible tools
Build a dual data strategy that covers Amazon Rufus and Buy for Me optimization in parallel
Maintain sovereignty over customer data — build a data foundation that prevents AI platforms from monopolizing your insights
Conclusion
The Open vs Walled Garden conflict is the structural dynamic of agentic commerce itself. Rather than one side winning completely, the most realistic future is coexistence — much like PCs and mobile coexist today — with an open protocol economy alongside Amazon's closed economy. What merchants need is not to bet on one side, but to build a flexible foundation that can serve both — and to build it fast.
