Agentic Commerce: How AI Agents Are Rewriting the Rules of Shopping in 2026

The End of Search-Then-Shop

For the past three decades, online shopping followed a predictable pattern. You searched — on Google, on Amazon, on a retailer's website. You evaluated results, clicked links, read product pages, and eventually made a purchase decision after navigating a maze of tabs, comparison windows, and checkout flows. The entire architecture of e-commerce was built around this sequential, user-driven model.

That model is now being dismantled, faster than most retail analysts predicted.

In 2026, a new paradigm called agentic commerce is emerging as the dominant trend in online retail. Instead of searching for products and navigating to purchase them yourself, AI agents do it for you — browsing, comparing, evaluating, and in some cases, completing the purchase entirely without manual intervention. The implications for retailers, consumers, and the trillion-dollar e-commerce industry are profound.

This piece examines the major players, their approaches, what agentic commerce means for everyday shoppers, and what the stakes are for the businesses that sell online.

Defining Agentic Commerce

Before examining the specific platforms, it is worth defining what "agentic" means in this context. An agentic AI system is one that takes autonomous action in the world to complete goals — not just answering questions, but performing multi-step tasks that involve browsing, decision-making, and executing transactions.

Traditional AI shopping assistance was reactive: you described a product, the AI returned information. Agentic shopping AI is proactive: you describe a need or set a preference, and the agent acts on your behalf through a sequence of steps — finding options, comparing them against your criteria, and potentially purchasing the best one.

The difference sounds subtle, but it restructures the entire consumer-retailer relationship. When an AI agent is the shopper, retailers must optimize their products for algorithmic discovery rather than human attention. And when an AI completes a purchase without the consumer ever visiting a product page, entire swaths of the traditional e-commerce interface become irrelevant.

OpenAI: Operator and Instant Checkout

OpenAI's journey into commerce began with its Operator feature, announced in early 2025, which allowed ChatGPT to browse websites and complete tasks on users' behalf. Shopping was an obvious application, but Operator's early attempts at navigating retail checkout flows were clunky and error-prone.

The more significant development came in late 2025 and early 2026 when OpenAI announced partnerships with major retailers including Target, Instacart, and DoorDash. These integrations introduced Instant Checkout — a purchasing system powered by Stripe that allows products to be bought directly within ChatGPT's interface without leaving the conversation.

The user experience is genuinely different from anything that existed before. You tell ChatGPT you need a birthday gift for your sister who likes cooking, under $75, shippable by Thursday. The agent queries its retail partners, surfaces three or four options with brief explanations tailored to your stated criteria, and presents a checkout button. Payment details saved in your profile mean the entire purchase can be completed in two taps.

OpenAI is careful to describe this as "assisted commerce" rather than fully autonomous purchasing. The human remains in the loop for the final confirmation. But the friction reduction is massive: the evaluation and discovery stages — historically the most time-consuming and cognitively demanding parts of online shopping — are handled by the AI.

Industry observers have noted that OpenAI's first foray into shopping through Operator was, by the company's own admission at a March 2026 CNBC interview, imperfect. "Our first crack at online shopping stumbled," an OpenAI representative acknowledged, citing issues with checkout flow navigation on non-partner sites. The Instant Checkout system with vetted retail partners is the course correction — a walled garden approach that prioritizes reliability over open-web browsing.

Perplexity: Buy with Pro and the Legal Battles

Perplexity AI was arguably the first major AI company to make a serious move into agentic shopping. The company launched its Buy with Pro feature for US subscribers in late 2024, allowing users to find products through Perplexity's search interface and purchase them without leaving the app.

Perplexity's approach is philosophically different from OpenAI's. Where OpenAI built direct retail partnerships, Perplexity initially attempted to enable purchasing across a broad range of e-commerce sites through its Comet shopping agent, which could navigate external checkout flows autonomously.

This triggered a significant legal confrontation. In early 2026, Amazon filed suit against Perplexity, alleging that the Comet shopping agent was making unauthorized purchases on Amazon's platform by automating checkout flows in ways that violated Amazon's terms of service. A US court issued a preliminary injunction blocking Comet from operating on Amazon's platform while the case proceeds. The legal outcome will likely set important precedents for what AI agents are permitted to do on retail platforms they do not have formal agreements with.

The Perplexity vs. Amazon lawsuit is a microcosm of a broader tension in agentic commerce: AI platforms want universal access to enable agents to shop anywhere, while retailers want to control how AI interacts with their platforms, pricing, and checkout flows. Expect this dispute to define the legal landscape of agentic commerce for years.

Despite the legal friction, Perplexity's Buy with Pro feature continues to function effectively on platforms that have not opposed it. The Snap to Shop feature — which lets users photograph a physical product and instantly find where to purchase it online — has become a standout feature with strong consumer adoption.

Amazon: Rufus, Alexa+, and the Ecosystem Defense

Amazon occupies the most unusual position in the agentic commerce landscape. It is simultaneously the world's largest e-commerce platform, a major AI company, and the most threatened incumbent in the space. AI shopping agents that route purchases to retail partners represent an existential threat to Amazon's core business model — search-to-purchase within Amazon's own ecosystem.

Amazon's response has been to build its own agentic commerce capabilities while refusing to participate in the open protocols (like OpenAI's UCP and Google's Commerce Protocol) that would allow competitors to route purchases through Amazon on their terms.

Amazon Rufus is the company's AI shopping assistant, integrated directly into the Amazon app and website. Rufus has scaled to 300 million users, according to figures cited in Q4 2025 earnings calls, and Amazon reports that Rufus-assisted product discoveries drive 60% higher conversion rates compared to standard search. The estimated $12 billion in incremental sales attributed to Rufus in 2025 validates the investment.

Rufus is now capable of understanding complex natural-language shopping requests — "Find me a noise-canceling headphone under $200 that works well for calls and has at least 30 hours battery life" — and returning curated results with comparative analysis. The experience within Amazon's ecosystem is genuinely impressive.

Alexa+, the enhanced version of Amazon's voice assistant with generative AI capabilities, extends this to voice-activated purchasing. Subscribers can complete purchases through voice commands with contextual awareness of their order history, Prime status, and delivery preferences.

The strategic calculus is clear: Amazon is building the best agentic shopping experience within its own ecosystem so that consumers have no reason to use third-party agents that might route purchases elsewhere. It is a defensive move, but also a genuine product investment.

Shop on Amazon with Rufus AI

Google: UCP and Gemini's Commerce Ambitions

Google's approach to agentic commerce centers on the Universal Commerce Protocol (UCP) — an open standard for AI agents to interact with e-commerce platforms. Google has signed UCP partnerships with Walmart, Target, Shopify, Etsy, and over 20 other retail partners, allowing Google's Gemini AI to access real-time inventory, pricing, and checkout APIs.

The Gemini Shopping experience integrates with Google's existing infrastructure: search, Maps, and Pay. When you ask Gemini to find and purchase a product, it can query UCP-participating retailers, compare options, and initiate checkout using your saved Google Pay information. The "Buy for Me" feature in Google's AI search results handles the full flow for supported products on supported platforms.

Google's UCP strategy is philosophically open compared to Amazon's closed approach. By creating a standard protocol, Google positions itself as the neutral infrastructure layer for agentic commerce, similar to how it positioned the Chrome browser and Android as open platforms that nonetheless feed Google's core ad and search business. Retailers who join UCP benefit from AI-driven discovery within Gemini's rapidly growing user base; Google benefits from transaction data and increased relevance as a commerce platform.

The integration with Google Search is significant. As of Q1 2026, Google has begun surfacing "Buy with AI" buttons directly in search results for qualifying products — a direct overlay on the traditional search-then-click-then-shop flow that disrupts the funnel at its earliest stage.

Perplexity's Snap to Shop: The Mobile Opportunity

One feature from the agentic commerce ecosystem that deserves separate attention is Perplexity's Snap to Shop functionality. Available on the Perplexity mobile app, it allows users to photograph any physical product — a friend's shoes, a coffee maker at someone's house, a product in a magazine — and receive immediate links to purchase that exact product or visually similar alternatives.

The technology combines visual recognition, product database matching, and real-time pricing lookups to bridge the physical-to-digital commerce gap that retailers have struggled with for years. Retailers have spent hundreds of millions trying to solve "reverse image search for shopping" through dedicated apps; Perplexity has made it a feature in a general-purpose AI assistant.

For consumers, Snap to Shop represents a fundamentally new impulse-purchase behavior. Instead of photographing something interesting and forgetting about it, the purchase pathway is immediate. This has implications for brand visibility in physical spaces, the value of product placement in influencer content, and the economics of visual advertising.

What Agentic Commerce Means for Online Retailers

For the businesses that sell online, the rise of AI shopping agents represents a more profound disruption than any previous digital commerce shift. Here is what the transition means in practical terms:

Discovery is no longer human. When AI agents are doing the shopping, SEO and product photography optimization matters less. What matters is whether your product data — specifications, pricing, availability, customer review signals — is structured in a way that AI systems can evaluate and prefer. Retailers who provide clean, machine-readable product data will win; those relying on visual persuasion and emotional marketing copy will find AI agents indifferent.

Brand loyalty weakens. AI agents optimize for stated criteria — price, specifications, shipping speed — not brand affinity. If a customer asks for "the best running shoe under $150 with high arch support," the agent will return the best match regardless of whether the consumer has historically bought Nike or Adidas. Brand loyalty was already weakening in e-commerce; AI agents accelerate that erosion.

The checkout page becomes obsolete. For AI-mediated purchases, the consumer may never visit a product page or checkout flow. Retailers will need to ensure their products are accessible through API-based commerce protocols (like UCP) and maintain real-time inventory and pricing feeds that AI agents can query directly.

McKinsey projects agentic commerce could reach $1 trillion in US retail revenue by 2030, with global projections in the $3–5 trillion range. These numbers reflect the fundamental restructuring underway: agentic AI becomes the primary interface for online commerce, with human direct browsing becoming the exception rather than the rule.

The Consumer Privacy and Trust Question

Agentic commerce involves AI systems with access to your purchase history, payment credentials, shipping preferences, and potentially even ongoing purchase mandates ("always buy the cheapest option in the preferred brand"). This level of data access creates privacy and security considerations that deserve serious scrutiny.

All four major platforms — OpenAI, Perplexity, Amazon, and Google — require explicit opt-in for shopping features and payment credentials. Transactions are completed with user confirmation rather than fully autonomously in most current implementations. But the architecture of agentic commerce creates vectors for data breaches, price manipulation (an agent trained on retailer-provided data may be biased toward certain retailers), and gradual erosion of consumer price sensitivity.

The regulatory environment is beginning to respond. The EU's AI Act, fully in force as of 2026, requires disclosure when AI agents complete commercial transactions on users' behalf. The US Federal Trade Commission has issued guidance on AI-mediated commerce disclosures. These are early-stage regulatory frameworks, and the legal landscape will evolve significantly as agentic commerce matures.

The Battle for the Middle Layer

The deeper strategic conflict in agentic commerce is a fight for what venture investors call "the middle layer" — the interface between consumers and retailers. Amazon built its dominance by owning the middle layer through its marketplace. Google competed by owning search — the entry point before Amazon. OpenAI, Perplexity, and now Google are betting that AI agents become the new middle layer, rendering both the search result page and the Amazon marketplace search box secondary to AI-mediated discovery.

For Amazon, this is an existential challenge. If shoppers stop searching on Amazon and instead ask ChatGPT or Gemini to find and purchase products, Amazon loses the discovery advantage that has made it the default starting point for product searches for over a decade. This explains the ferocity of Amazon's legal response to Perplexity's Comet agent and the urgency behind Amazon's own Rufus and Alexa+ investments.

The outcome of this battle will determine the distribution of economic power in digital commerce for the next decade. Whether open protocols like UCP create a genuinely level playing field, or whether closed ecosystems like Amazon's maintain their dominance, will be shaped by regulatory decisions, consumer behavior patterns, and the technical quality of the competing AI shopping experiences.

What You Should Do as a Consumer in 2026

For everyday shoppers, agentic commerce offers real benefits today — if you are willing to engage with the early technology:

For time-sensitive purchases: ChatGPT's Instant Checkout on supported retailers is genuinely faster for products where quality is commoditized. Office supplies, household basics, and standard tech accessories are good candidates for AI-assisted purchasing.

For research-heavy purchases: Perplexity and Google's Gemini Shopping provide strong comparative analysis for complex purchases — electronics, appliances, and anything where specifications matter. The AI's ability to filter and compare across multiple criteria simultaneously is faster than manual research.

For deal-finding: Amazon Rufus's integration with price history data and Prime member pricing makes it effective for identifying when a product has dropped to a historically good price. The "track this product" feature with price alerts in the Rufus interface is practical for non-urgent purchases.

Keep your own judgment: AI agents optimize for the criteria you explicitly state. Implicit factors — brand trust built on long experience, aesthetic preferences that are hard to articulate, intuitions about product quality — are currently poorly handled by agentic systems. For significant or emotionally resonant purchases, direct browsing and human evaluation still has clear advantages.

The age of agentic commerce is beginning, not ending. The systems are imperfect, the legal frameworks are incomplete, and the consumer habits are still forming. But the direction is unmistakable. Within five years, a meaningful fraction of all online purchases will be initiated by AI agents rather than human browsers. The retail industry, the advertising industry, and consumer behavior patterns will all be transformed in the process.

The question is not whether AI will reshape how we shop. It is which AI platform will own the experience when it does.