AI Agents vs. Marketplaces: Who Controls E-commerce?

Executive Summary:

• The emergence of autonomous AI shopping agents, such as Perplexity’s Comet and OpenAI’s Atlas, is rapidly redefining product discovery and purchasing in e-commerce.

• Leading marketplaces like Amazon and eBay are actively implementing legal and policy measures to restrict the uncontrolled operation of these AI agents on their platforms.

• Amazon initiated a lawsuit against Perplexity in November 2025, alleging that Comet violates terms of service by ‘masquerading as a human’ and executing orders.

• eBay updated its user agreement in January 2026, prohibiting unsanctioned ‘buy-for-me’ agents and end-to-end LLM-driven checkouts to protect against fraud and maintain control.

• Marketplaces’ primary concerns include protecting their substantial advertising revenue (Amazon reported $47 billion in ad services in the first nine months of 2025), maintaining control over first-party shopper data, and ensuring transactional integrity.

• The unfolding conflict is expected to lead to a compromise involving negotiated API access and potential commercial arrangements for AI agents, impacting the future of retail and digital marketing.

The rapid evolution of autonomous AI shopping agents, exemplified by platforms like Perplexity’s Comet and OpenAI’s Atlas, is ushering in a new era of agentic commerce. This paradigm shift allows artificial intelligence to independently search, compare, and initiate purchases with minimal human intervention, promising consumers unprecedented efficiency. However, this transformative technology presents a profound challenge to established e-commerce conventions, particularly for dominant online marketplaces. Giants such as Amazon and eBay are reacting decisively, deploying legal actions and stringent policy updates to curtail the activities of these independent AI agents, citing security and user experience concerns while fundamentally seeking to protect their core business models. This unfolding conflict highlights a critical juncture for the digital economy, with significant implications for product discovery, advertising revenue, and the overall control of the online retail ecosystem, as reported by Practical Ecommerce on January 29, 2026.

Analysis

The clash between AI innovation and marketplace control has rapidly escalated into both legal and policy arenas. In November 2025, Amazon initiated a lawsuit against AI-first browser developer Perplexity, accusing its Comet browser of ‘masquerading as a human’ and executing orders on Amazon accounts in violation of the marketplace’s terms of service and computer fraud statutes. Amazon’s official stance, per a company statement, asserts that third-party bots must operate transparently and only with explicit platform permission. Perplexity, in its counter-argument, posits that Comet acts as an authorized proxy for human users, storing credentials locally for enhanced security, and suggests Amazon’s litigation is a strategic maneuver to safeguard its substantial ad-driven revenue model and maintain absolute control over the entire shopping journey. Further reinforcing this trend, eBay took proactive measures in January 2026, revising its user agreement to explicitly forbid, without prior approval, ‘buy-for-me’ agents and any end-to-end Large Language Model (LLM)-driven checkout processes. While eBay frames this policy as a safeguard against auction manipulation, fraud, and mistaken orders, it strategically leaves room for ‘formally sanctioned’ shopping agents, signaling a clear intent to dictate the terms of AI integration and prevent an uncontrolled erosion of its platform dominance. The aggressive responses from Amazon and eBay are not merely about security; they reveal a deep concern for the fundamental pillars of their marketplace business models. These platforms thrive on aggregating product listings and centralizing the shopping experience, which directly underpins their revenue generation. The rise of agentic commerce directly threatens this established order. A chief concern for these marketplaces is their colossal advertising revenue. For instance, Amazon reported a staggering $47 billion in ‘advertising services’ revenue during the first nine months of 2025 alone, according to its Q3 2025 SEC filing. This figure underscores Amazon’s dual role as both a product marketplace and a powerful digital publisher, leveraging sponsored listings, recommendation units, and paid placements deeply integrated into its search results and category pages. Autonomous AI agents, by design, are engineered to bypass these commercial placements, identifying and proceeding directly to purchase decisions, thus circumventing the very advertising mechanisms that generate billions for the platforms. Beyond advertising, the proliferation of AI shopping agents poses significant risks to marketplaces’ control over first-party shopper data and the integrity of transactions. Marketplaces like Amazon meticulously collect and analyze vast amounts of behavioral data: search queries, product views, abandoned carts, and purchase histories. This data is crucial for feeding proprietary ranking algorithms, refining recommendation systems, and personalizing the user experience, ultimately driving further engagement and sales. When an external AI agent conducts product comparisons and decision-making outside the marketplace’s direct observation, this invaluable data stream is severely diminished, leaving the platform with only the final purchase record. Furthermore, while AI agents like Perplexity’s Comet can complete transactions via the marketplace’s checkout, this creates concerns for the platform. Marketplaces lose the ability to guarantee the propriety of the transaction, leading to potential issues with erroneous orders, increased customer service inquiries, or higher return rates. Crucially, the opportunity for upselling and cross-selling, a significant revenue driver traditionally facilitated by human interaction or carefully designed UI flows, becomes virtually impossible when the human shopper never actively engages with these prompts.

Why it matters

This evolving landscape signals a pivotal moment for e-commerce and digital marketing. The actions of Amazon and eBay underscore that the control over product discovery and the flow of transactions is now a central battleground. For businesses engaged in Generative Engine Optimization (GEO) and technical-editorial SEO, this shift means understanding that future discovery may increasingly occur through AI models rather than traditional search engine results or marketplace browsing. The impact on traditional SEO and digital advertising could be profound, as AI agents prioritize efficiency and relevance over sponsored placements. The current standoff between marketplaces and AI developers, however, will likely culminate in a form of compromise. This resolution will likely involve marketplaces granting API access to approved AI agents, subject to stringent rate limits, robust identity verification protocols, and potentially novel commercial arrangements. These arrangements could mirror existing affiliate programs, where AI bots earn commissions for facilitating sales. For small-to-medium ecommerce businesses, adapting to this new reality is paramount. Such approved AI agent-marketplace relationships could become a vital new channel for product distribution, offering a direct route for their products to be discovered and purchased via AI assistants integrated into platforms like ChatGPT or Perplexity. This necessitates a proactive strategy to optimize product data and integrate with emerging AI commerce frameworks, ensuring visibility and transactional readiness in an increasingly automated retail environment.