While Google, Stripe, and Shopify were building agentic commerce protocols throughout January 2026, Amazon quietly proved the concept already works. Rufus, Amazon’s AI shopping assistant, generated nearly $12 billion in incremental annualized sales during 2025. More than 300 million customers used it. Shoppers who interact with Rufus convert at 60% higher rates than those who do not. And since January 2026, Rufus can autonomously purchase products for you when prices drop to a target you set.
That makes Rufus the first agentic shopping agent operating at genuine scale. Not a demo, not a pilot, not a protocol spec waiting for adoption. A working system that handles product discovery, personalized recommendations, price tracking, and autonomous purchasing for hundreds of millions of people.
From Q&A Bot to Autonomous Buyer: How Rufus Evolved
Rufus launched in February 2024 as a conversational search tool inside the Amazon app. You could ask it questions like “what’s the best budget espresso machine?” and get answers drawn from product listings, reviews, and Q&A sections. Useful, but not fundamentally different from a search bar with better natural language processing.
Throughout 2025, Amazon deployed more than 50 technical upgrades that turned it into something qualitatively different. The evolution happened in three phases.
Phase 1: Memory and Personalization
Rufus gained persistent memory across sessions. If you asked about running shoes last month, it remembers your size, brand preferences, and budget range. This is not session-based context (which chatbots have done for years). It is cross-session personalization that builds a buyer profile over time. Amazon calls this “shopping DNA,” and it is built on the same purchase history, browsing patterns, and review interactions that power Amazon’s recommendation engine, now exposed through a conversational interface.
Phase 2: Price Intelligence
The second phase added price tracking that goes beyond simple alerts. Rufus can analyze historical pricing patterns, identify deals in context (“this is the lowest this item has been in 90 days”), and proactively surface pricing opportunities. Unlike third-party tools like CamelCamelCamel, Rufus has direct access to Amazon’s internal pricing data, including upcoming Lightning Deals and Prime-exclusive prices that external trackers cannot see.
Phase 3: Auto Buy (Autonomous Purchasing)
This is where Rufus crossed from assistant to agent. Launched in January 2026, the Auto Buy feature lets Prime members set a target price for any Fulfilled-by-Amazon item. Rufus then monitors that price every 30 minutes and completes the purchase autonomously when the price hits the target, using your default payment method and shipping address. You get a notification with a 24-hour cancellation window.
The key distinction: Auto Buy is not “notify me when the price drops.” It is “buy it for me when the price drops.” That shift from notification to execution is what separates an assistant from an agent.
Amazon’s Auto Buy help page lists the constraints: Prime members only, Fulfilled-by-Amazon items only, default 6-month monitoring window. Customers using Auto Buy save an average of 20% per purchase, according to Amazon’s own metrics.
The Numbers: $12 Billion and 60% Higher Conversion
Amazon does not break out Rufus revenue in its earnings reports, but two data points tell the story.
The first: Fortune reported in November 2025 that Rufus was on pace for $10 billion in incremental annualized sales after Q3. By year-end, that figure hit nearly $12 billion, suggesting accelerated adoption during the holiday season.
The second: customers who engage with Rufus during a shopping session are 60% more likely to make a purchase. That is a conversion lift that any e-commerce platform would consider transformative. For context, Amazon’s overall conversion rate is already among the highest in e-commerce at roughly 10-15% for Prime members. A 60% lift on top of that is significant.
What drove the growth was not a single feature but the compound effect of the 50+ upgrades. Rufus went from answering 2.5% of shopping queries in mid-2024 to being present in a meaningful share of the 300+ million user sessions by late 2025. Each upgrade, better product comparison, visual search, integration with Amazon’s advertising system, made it more useful, which drove more usage, which generated more training data, which made it more useful again.
For sellers, the impact is mixed. Rufus-influenced purchases tend to favor items with strong reviews and clear product descriptions, since Rufus parses these for its recommendations. Products with sparse or poor listing content become effectively invisible to Rufus, which now mediates a growing share of product discovery on Amazon.
Buy for Me: Shopping Beyond Amazon
Auto Buy handles purchases within Amazon. But Rufus also gained a “Buy for Me” feature that extends its reach beyond Amazon’s own marketplace. When you search for branded products, Rufus can browse tens of millions of items on third-party websites, compare them against Amazon listings, and facilitate purchases on those external sites using encrypted payment and shipping information.
This is Amazon’s response to what CNBC called “Amazon’s AI agent dilemma”: fight third-party AI shopping agents or join them. Amazon chose a third option. Build its own agent that shops everywhere, keeping the customer relationship while expanding beyond Amazon’s catalog. The strategic implication is clear: if Rufus becomes the default shopping interface for 300+ million customers, it controls product discovery across the entire internet, not just Amazon’s marketplace.
There is an important nuance. Unlike Auto Buy, the Buy for Me feature still requires you to approve and execute the purchase. It handles browsing, comparing, and cart building, but the final click stays with the human. Amazon likely structured it this way to limit liability for purchases on third-party sites where return policies and fulfillment standards vary.
Rufus vs. the Competition: Who Wins the Shopping Agent War
Rufus is not the only AI shopping agent, but it has a structural advantage that competitors cannot easily replicate.
The Data Moat
Google Gemini Shopping has access to product feeds from millions of merchants. ChatGPT Shopping has deals with Target, Instacart, and DoorDash. Perplexity Shopping has PayPal integration. But none of them have what Amazon has: 25+ years of purchase history data from hundreds of millions of customers, combined with real-time inventory, internal pricing data, and the world’s largest fulfillment network.
When Rufus recommends a product, it is drawing on what you actually bought (not just browsed), what similar buyers purchased, return rates, review sentiment, and real shipping times from Amazon’s logistics network. This is a fundamentally richer signal than what any external agent can access by scraping product feeds.
The Conversion Gap
Retail Technology Innovation Hub’s analysis puts this in context: the battle for customer intent starts earlier than the shopping session. Google controls search intent. ChatGPT captures conversational intent. But Amazon controls transactional intent, the moment when someone is ready to spend money. Rufus intercepts that intent with a more complete understanding of the customer than any competitor can build from public data alone.
The Anti-Amazon Alliance
Competitors recognize this advantage. Google’s Universal Commerce Protocol was partly designed to let third-party agents access merchant catalogs that compete with Amazon. OpenAI’s partnerships with Target, Walmart, and Instacart give ChatGPT direct access to purchase flows that bypass Amazon entirely. Perplexity’s agent-powered browser caused legal friction with Amazon over how it accessed Amazon’s product pages.
The emerging picture: an “anti-Amazon alliance” where competing platforms build interconnected agentic commerce infrastructure specifically designed to prevent Amazon from capturing the entire AI shopping layer. Whether that strategy works depends on whether convenience (one agent, one ecosystem, one-click purchasing) beats choice (multiple agents, open protocols, cross-retailer comparison).
What This Means for Sellers and Retailers
The rise of Rufus changes the optimization game for anyone selling on or competing with Amazon.
Amazon sellers need to optimize for Rufus, not just search. Traditional Amazon SEO focused on keyword-stuffed titles and backend search terms. Rufus parses product descriptions, Q&A sections, and review content differently than keyword-based search. Products need clear, structured descriptions that answer the questions Rufus asks on behalf of buyers: “Is this compatible with X?”, “How does this compare to Y?”, “What are the common complaints?”
Advertising is merging with agent recommendations. Amazon has begun integrating sponsored content within Rufus conversations. When Rufus recommends products, some of those recommendations are paid placements. This is the agentic equivalent of Google’s paid search results, and it represents a new advertising channel with potentially higher conversion rates since Rufus users already convert 60% better.
Retailers outside Amazon face a discovery problem. If Rufus becomes the default product discovery interface for 300+ million shoppers, products not on Amazon become harder to discover. The Buy for Me feature partially addresses this by letting Rufus shop external sites, but Amazon controls which external products surface and how prominently. For retailers, joining Google’s UCP or Stripe’s Agentic Commerce Suite is no longer optional: it is the only way to ensure AI agents can find and recommend your products.
The Trust Equation: Would You Let an AI Spend Your Money
The 60% conversion lift is real, but it masks a harder question. Only 12% of consumers currently trust AI to make purchases autonomously, even though 70% say they are at least somewhat comfortable with AI shopping assistance. Amazon’s Auto Buy threads this needle by offering a specific, constrained form of autonomy: you set the price, the item, and the payment method. The agent just watches and executes.
That model, human sets the parameters and the agent operates within tight guardrails, is probably how agentic commerce scales. Not fully autonomous agents making open-ended purchasing decisions, but specialized agents that execute well-defined tasks within clear boundaries. Amazon designed Auto Buy this way deliberately: the 24-hour cancellation window, the restriction to Prime and FBA items, the default payment method requirement. Each constraint reduces the trust gap between “I’d use an AI to find products” and “I’d let an AI spend my money.”
The question for 2026 is whether this constrained model expands or whether consumer trust stays anchored at low levels. Amazon’s data will be the best signal: if Auto Buy adoption grows significantly through 2026, it validates the guardrailed autonomy approach. If it plateaus, it suggests that the trust barrier is real and that agentic commerce will remain primarily a discovery and comparison tool rather than an autonomous purchasing system.
Frequently Asked Questions
What is Amazon Rufus?
Amazon Rufus is an AI-powered shopping assistant built into the Amazon app. It uses generative AI trained on Amazon’s product catalog, customer reviews, and Q&A data to answer shopping questions, compare products, track prices, and autonomously purchase items through its Auto Buy feature. As of late 2025, more than 300 million customers used Rufus.
How does Amazon Auto Buy work?
Auto Buy lets Amazon Prime members set a target price for any Fulfilled-by-Amazon item. Rufus monitors the price every 30 minutes and automatically completes the purchase when the price hits the target, using your default payment method and shipping address. You receive a notification with a 24-hour cancellation window. Customers using Auto Buy save an average of 20% per purchase.
How much revenue did Amazon Rufus generate?
Amazon Rufus generated nearly $12 billion in incremental annualized sales during 2025, up from the $10 billion annualized pace reported after Q3 2025. Customers who engage with Rufus during shopping sessions convert at 60% higher rates than those who do not use it.
What is the difference between Rufus Buy for Me and Auto Buy?
Auto Buy is fully autonomous: you set a target price and Rufus buys the item when the price drops, without further input. Buy for Me extends Rufus to third-party websites outside Amazon, letting it browse and compare items across the web, but still requires you to approve and execute the final purchase. Auto Buy works only on Amazon with Fulfilled-by-Amazon items.
How does Amazon Rufus compare to Google Gemini Shopping and ChatGPT Shopping?
Rufus has a structural advantage: 25+ years of Amazon purchase history, real-time internal pricing, and direct integration with Amazon’s fulfillment network. Google Gemini Shopping offers checkout via the Universal Commerce Protocol across multiple retailers. ChatGPT Shopping has partnerships with Target, Walmart, and others. Rufus has the deepest data on individual shoppers, while Google and ChatGPT offer broader cross-retailer comparison.
We cover AI agents from protocols to production. Subscribe for practical guides every week.