Amazon Replaced Rufus with Alexa for Shopping, an AI That Buys for You

More than 300 million people used Amazon's Rufus chatbot in 2025. They asked it which dishwasher was best for a large family, whether a $120 Bluetooth speaker was worth the money, what laptop a college student needed for graphic design work. Rufus answered those questions decently, sometimes well. What it never did was finish the job.

That changes today. Amazon retired Rufus on Wednesday and replaced it with a fundamentally different product: Alexa for Shopping, an AI assistant that doesn't just answer questions about what to buy. It monitors prices over time, schedules purchases automatically, builds personalized shopping guides, and buys things for you on websites Amazon doesn't even control. That last part is the shift that matters.

Amazon has spent two years telling investors and analysts that it is serious about AI. Rufus was the first public proof point, a chatbot built into the Amazon app that could hold a product research conversation and surface structured comparisons. Rufus worked. People used it. But it was still, at its core, a search engine dressed in conversational clothes. It surfaced information and stopped there.

Alexa for Shopping is built on a different premise. It's designed to take action, not just provide answers. The shift from answering questions to completing tasks is exactly what AI companies mean when they talk about agentic AI, and Amazon is betting that shopping is the domain where consumers will actually trust an AI to work independently on their behalf.

What Alexa for Shopping Actually Does

Alexa for Shopping launched Wednesday across the Amazon mobile app, Amazon.com, and Echo Show smart displays. No Prime membership is required, which means Amazon is opening the product to its entire customer base rather than treating it as a premium feature tier. Access is a tap on a stylized lowercase "a" icon in the corner of Amazon's app or website. On the desktop, the assistant opens in a dedicated panel alongside your existing search results.

Start with the basics. Type a question into Amazon's search bar and you get a guided response instead of a raw list of results. Ask what you need for a week-long camping trip with a dog in the Pacific Northwest, and Alexa for Shopping will recommend specific product categories with explanations of why each matters, not just a list of links. Ask for a laptop for a design student, and it generates a comparison table of five specific models with a year of price history for each.

The year-long price history is one of the most immediately useful features. Amazon has historically made it difficult to track how much a product has actually cost over time, partly because third-party sellers adjust prices frequently and inflate sale discounts. Having that historical data surface automatically makes it significantly harder to be fooled by a fake discount. Price history is displayed per product automatically, and the assistant factors it into its recommendations without the user having to ask.

Beyond individual product lookups, Alexa for Shopping can build custom shopping guides for major purchase categories. A guide to home theater setups, for instance. A breakdown of standing desk options by price range and feature set. A roundup of infant car seat choices ranked by safety ratings and price. These are personalized based on your purchase history, meaning a parent who has bought three child-safety products in the past year will get recommendations calibrated to that household rather than a generic consumer baseline.

The autonomous purchasing feature Amazon calls "Buy for Me" is where Alexa for Shopping becomes genuinely different from every prior version of AI-assisted shopping. Instead of finding a product and completing the purchase yourself, you can ask Alexa for Shopping to monitor a product and buy it when conditions are met: when the price drops below a threshold you set, when a specific date arrives, or when you're running low on a routine item. That's already a significant capability. But the more ambitious version is what makes this a platform-level shift: Alexa for Shopping can complete purchases on third-party websites outside Amazon entirely, acting on your behalf without requiring you to do anything.

That's a significant departure from how Amazon has historically operated. Amazon built its entire business model around being the destination for purchases. The inclusion of cross-retailer autonomous purchasing suggests Amazon is willing to sacrifice some direct transaction revenue in exchange for becoming the AI layer through which people shop across the entire web. If Alexa for Shopping becomes the assistant people use to shop everywhere, Amazon becomes the dominant intermediary in e-commerce regardless of where the transaction settles. The data advantage alone, knowing what people bought across retailers, what prices triggered purchases, what categories they shopped across multiple sites, is worth more to Amazon's advertising and recommendation business than the margin on any individual transaction.

One of the more subtle capabilities is how the product uses contextual memory across Amazon's entire device ecosystem. The assistant pulls data from multiple sources simultaneously: your product searches on Amazon.com, items you've purchased, pages you've browsed, and conversations you've had with Alexa on Echo devices in your home. These threads get woven together into a unified model of your household and your needs.

If you had a conversation with your Echo Show last month about a science fair project your child was working on, you can later ask Alexa for Shopping on your phone to suggest supplies for that same project, and it will retrieve that context. You don't need to re-explain the project. It already knows. The same memory applies to products you own. Ask Alexa for Shopping for a compatible cleaning tablet for your dishwasher and it already knows which dishwasher model you have, because you bought it through Amazon two years ago. Ask it to troubleshoot an error code on your refrigerator and it pulls the model from your purchase history. The assistant behaves less like a search engine and more like a household manager with a running memory of what you own and what you need.

This cross-device memory is what separates a genuinely agentic assistant from a smarter search bar. Search engines retrieve. Agents remember and act in context. The distinction sounds abstract until you use a system where the AI already knows your household setup, your purchase patterns, and your standing preferences without you having to re-explain them every time.

Amazon simultaneously launched a new visual shopping experience for Echo Show 15 and Echo Show 21 devices, extending Alexa for Shopping's capabilities to the large-screen displays many households have mounted in their kitchens or common areas. Previously, Echo Show's shopping functionality was primarily voice-driven with limited visual support. You could ask Alexa to reorder something and it would do it, but browsing and comparing products on the display was awkward and rarely used. The new experience presents full product pages, side-by-side comparisons, and searchable results directly on the display, with touch interaction on compatible models. The kitchen context is particularly relevant: shopping decisions made while cooking, running low on an ingredient, or noticing a household item needs replacement happen at home in real time. A large-screen display that's already logged into your Amazon account, tracking your shopping history, and watching for price drops is more useful in that context than pulling out a phone for the same task.

For Amazon, this memory layer also solves a longstanding product discovery problem. The company's recommendation engine has historically been pattern-matching at scale: customers who bought X also bought Y. Contextual memory allows for something more specific. A customer who owns six specific items, has a child entering middle school, consistently buys within a particular price range in this category, and has previously expressed a preference for a specific brand gets recommendations that are qualitatively different from what a pattern-match engine can produce. That quality gap is where Alexa for Shopping expects to win.

Why Amazon Retired Rufus After 300 Million Users

Rufus launched in 2024 as Amazon's first serious attempt at conversational AI in shopping. By any conventional measure, it succeeded. More than 300 million customers used it in 2025 alone. It demonstrably improved product research for many of them. It was a better experience than keyword search for complex purchase decisions. Rufus was particularly useful for categories where the specification language is confusing, comparing camera lenses, picking a mattress, navigating the labyrinthine world of home networking equipment, and it handled those use cases well.

But Rufus had structural limitations that weren't fixable with incremental improvements.

First, Rufus was a standalone product. It existed separately from Alexa+, Amazon's generative AI upgrade to its classic smart-home assistant. That meant users had two different AI interfaces within Amazon's ecosystem with no shared memory or coordinated capabilities. Rufus knew your shopping history. Alexa+ knew your home routines and connected device data. Neither knew what the other knew, which created obvious gaps when a shopping decision involved context that lived in the smart-home layer.

Second, Rufus couldn't act. It could research, compare, and recommend. It couldn't buy, monitor, or schedule. Every interaction ended with Rufus handing the task back to the user for completion. That handoff is exactly what agentic AI is designed to eliminate. Handing someone a recommendation and saying "now you go complete the purchase" is still search; it just happens to be conversational search.

Third, maintaining two separate products with separate branding created user confusion and split Amazon's AI development resources. Customers who used Rufus for product research then switched to Alexa+ for their Echo Show shopping queries were navigating two products that should have been one from the start. Building and improving Alexa for Shopping as a single integrated product is more efficient than maintaining parallel systems.

Retiring Rufus is the right call for the same reason the old separation became untenable: Rufus was built for a different era of AI capability. What made sense in 2024, when conversational product research was novel, doesn't scale into 2026, when the expectation is an AI that completes the shopping task rather than assists with part of it. Alexa for Shopping inherits everything Rufus knew and adds everything Rufus couldn't do.

The user data advantage compounds here. The richer the data model Amazon builds on each shopper through Alexa for Shopping, the harder it becomes for any competitor to replicate the personalization quality without equivalent data access. Users who opt into the "Buy for Me" feature are giving Amazon visibility into purchasing decisions and transaction data from websites outside Amazon.com, not just their Amazon history. For many consumers, the tradeoff is visible and acceptable: more useful assistance in exchange for more active use of their data. Those who've been deliberate about limiting what Amazon knows should read the terms carefully before enabling autonomous cross-retailer purchasing.

Amazon's Agentic Shopping Bet Against Meta and Google

Amazon isn't the only company trying to build the dominant AI shopping agent. The competitive dynamics around agentic commerce are accelerating, and the companies that establish category leadership now will have a structural advantage that compounds over time.

Meta is building Hatch, an AI agent for Instagram Shopping that takes a different entry point, social discovery rather than search intent. When you're scrolling Instagram and see a product you want, Meta's agent closes the loop without interrupting the browsing experience. The theory is that social discovery is where purchasing intent forms for many categories, particularly fashion, home goods, and lifestyle products, and meeting that intent at the discovery moment is worth more than competing for the purchase search.

Google's approach centers on its Shopping Graph, the world's largest product knowledge database, combined with Gemini integration in Search. When someone searches for a product on Google, the Shopping Graph has billions of product listings, pricing data, and user reviews to draw on. Google's agentic commerce layer sits on top of that existing data advantage.

Amazon's theory is different from both. Amazon's position starts with the moment of purchase intent: you go to Amazon when you already know you want something. Alexa for Shopping meets consumers at that moment and tries to complete the purchase as efficiently as possible, while extending the product's reach to the discovery phase through personalized guides and recommendations. These aren't minor tactical differences. They're distinct theories about where agentic commerce will actually become behavior that consumers adopt at scale. Amazon's version has the structural advantage of starting with hundreds of millions of existing customers who already go to Amazon to buy things.

For third-party sellers on Amazon, who account for roughly 60% of all units sold on the platform, Alexa for Shopping introduces dynamics that don't fully resolve in their favor. On the positive side, Alexa for Shopping's recommendation layer is built to surface relevant products across Amazon's entire catalog, not just Amazon's own brands. If a third-party seller's product has strong reviews, competitive pricing, and matches a user's stated needs, Alexa for Shopping should recommend it. The system is incentivized to produce useful recommendations because useful recommendations drive purchases, and purchases are what Amazon monetizes.

But the "Buy for Me" cross-retailer purchasing feature introduces a more complicated question. Amazon has not disclosed the decision logic it uses when choosing between Amazon-listed products and products available on competing sites. When the same item is available at the same price on Amazon and on a competing retailer, how does the assistant decide where to purchase? That question matters enormously to marketplace sellers. The price monitoring and automation features also accelerate pricing pressure in an already competitive environment. If Alexa for Shopping can monitor prices and wait for a target threshold before purchasing, sellers face a version of the race-to-the-bottom dynamic that already characterizes Amazon's third-party environment, but now driven by AI automation rather than manual comparison shopping. A seller who maintains a higher price while a competitor undercuts them by a dollar will lose the Alexa for Shopping purchase automatically, without the consumer making an active decision.

For a current overview of how AI agents are being deployed across consumer and enterprise applications, AIntelligenceHub's Agent Tools Comparison covers the evolving landscape.

Alexa for Shopping begins rolling out to all U.S. customers starting this week across the Amazon mobile app and website, available immediately and free of charge, with no Prime subscription required. The Echo Show visual shopping experience is live on the 15 and 21 models. Amazon hasn't committed to a specific timeline for broader Echo Show expansion, but the infrastructure it's building, a unified agentic layer across its entire device and service ecosystem, suggests this is a foundation for a much larger platform play rather than a feature release.

The most telling signal is what Amazon chose to retire. Rufus had 300 million users. Those users didn't go anywhere. Amazon is now betting that those same people, given a more capable and agentic assistant, will hand over more of their shopping behavior to an AI that actively acts on their behalf rather than just advising them. The size of that bet makes clear that this isn't an incremental product update. It's Amazon's statement about what shopping looks like from here.