The Amazon Shopping Interface Brands Know Is Being Replaced

In his annual shareholder letter published April 9, Amazon CEO Andy Jassy wrote that the company’s retail business is approaching $600 billion in topline revenue while roughly 80 percent of global retail sales still happens in physical stores. “That will change,” he wrote, and moved on. Earlier in the same letter, he was more specific about what Amazon intends to change closer to home: the consumer shopping experience itself.

“The temptation is to just add a little AI to the existing experience,” Jassy wrote. The harder and more valuable path, in his framing, is “reimagining your experiences from a clean sheet of paper, assuming you were building with the new technology available.” He named Amazon’s retail consumer experience as one of the candidates for exactly that treatment, alongside Alexa, which the company has already rebuilt once under live operating conditions.

For brands and CPG companies whose discovery, conversion, and advertising returns run through Amazon’s current architecture, parts of the new architecture are already running.

Rufus and the Mechanics of the New Discovery Layer

Amazon’s Rufus generative shopping assistant is the most visible expression of what a rebuilt shopping experience looks like. More than 300 million customers used Rufus in 2025, and customers who engage with it are about 60 percent more likely to complete a purchase, Amazon disclosed in its Q4 2025 earnings call. As of the Q3 2025 earnings call, Rufus was tracking toward over $10 billion in incremental annualized sales. Rufus usage grew 127 percent between Prime Week and Black Friday 2025, according to Similarweb data.

Rufus does not operate on keyword match. It draws on listing content, customer reviews, Q&A sections, A+ content, and browsing behavior to construct recommendations in response to natural-language queries. A shopper asking “what’s the best protein powder for someone who doesn’t like chalky textures” gets a curated answer, not a ranked list of keyword-matched results. What appears in that answer is determined by how completely and accurately a product’s content answers the question, not by bid or search term coverage.

Research from Mars United, based on analysis of more than 1,000 products, found that Rufus recommendations cluster around items with a minimum four-star rating and averaging close to 9,000 reviews per product. That is one study with a specific methodology, and Amazon has not published its own thresholds, but the directional finding is consistent with Amazon’s product listing guidance, which treats review hygiene and content completeness as inputs into AI recommendation eligibility. Products below those thresholds were largely absent from Rufus outputs in the Mars United analysis.

The practical exposure differs by brand type. Large CPG companies with established review volumes and dedicated content operations can treat Rufus optimization as an extension of existing digital shelf work. For mid-market brands, DTC companies scaling into Amazon wholesale, or category entrants with limited review history, the review and content requirements Rufus appears to apply represent a structural visibility gap that keyword investment alone cannot close. Amazon has introduced Sponsored Prompts as a paid placement layer inside Rufus interactions, which offers one route to visibility for brands that don’t yet meet the organic criteria. That format is early; the attribution and optimization reporting for Sponsored Prompts does not yet match the depth available for standard sponsored products.

The Alexa+ rebuild illustrates both the organizational method and the commercial upside Amazon expects from these changes. Amazon rewired Alexa’s reasoning layer, service routing, and knowledge base while the existing system continued serving 600 million active endpoints. The results Jassy cited in the letter include purchase completion on devices tripling and music streaming up 25 percent. The retail shopping experience is next in the same sequence.

The Grocery Build as Proof of Method

Jassy’s confidence in the clean-sheet approach is grounded in a recent completed example. Amazon’s grocery business reached over $150 billion in gross sales in 2025, making it the second-largest grocer in the United States, per the shareholder letter. That position was not planned in a straight line. Amazon started with non-perishable shelf-stable goods, acquired Whole Foods in 2017, launched and partially contracted Amazon Fresh physical stores, and eventually found its decisive move in integrating perishables into the Same-Day Delivery network in early 2025. Since that integration, perishable sales have grown more than 40 times, and fresh food now accounts for nine of the ten most-ordered same-day delivery items in regions where the service is available, across more than 2,300 cities and towns.

The delivery infrastructure under that grocery business is also running parallel bets rather than a single path. Over 85 Same-Day Fulfillment Centers hold the top 90,000 SKUs and have enabled more than 500 million same-day units delivered in 2026 thus far. Prime Air is on track to serve communities covering 30 million customers by year-end, with a target of half a billion drone deliveries by the end of the decade. Amazon Now, piloted in India and the UAE, is seeing 25 percent month-over-month order growth in India across more than 360 micro-fulfillment centers, with Prime members tripling their shopping frequency after adopting the service. Jassy frames these not as competing programs but as complements: drones use Same-Day Fulfillment Centers as launch infrastructure, and Amazon Now covers a faster delivery tier with narrower selection.

Amazon has also committed over $4 billion to a rural delivery network expansion covering more than 13,000 zip codes across 1.2 million square miles, with monthly same-day customers in rural areas nearly doubling in 2025 year over year. The letter projects the network will handle over one billion additional packages annually once complete.

Advertising Revenue and the Transition Period

Amazon’s advertising business added over $12 billion in incremental revenue in 2025, reaching $21.3 billion in Q4 at 22 percent year-over-year growth, per the Q4 2025 earnings call. eMarketer projects U.S. retail media spending will reach $69.33 billion in 2026, up nearly 18 percent from $58.79 billion in 2025, with Amazon Ads and Walmart Connect together capturing 89.5 percent of incremental spending. Rufus adds a new ad inventory surface to that ecosystem through Sponsored Prompts inside conversational shopping sessions, creating a paid visibility layer that did not exist in Amazon’s advertising architecture two years ago.

Retail media buyers managing Amazon budgets are currently operating across two discovery mechanics simultaneously. Traditional sponsored products and display formats run on auction and attribution systems with established optimization discipline. Rufus-native formats generate early data but lack the reporting depth to support comparable rigor. Organic Rufus visibility and paid Rufus placement are influenced by different inputs: a brand can buy its way into a Sponsored Prompt but cannot buy its way past a rating threshold or a content gap that Rufus uses to filter organic recommendations. The implication for retail media teams is that managing both surfaces as a single media problem will produce inconsistent results.

The AI discovery surface extends beyond Amazon. eMarketer’s January 2026 AI commerce report found that 39 percent of U.S. consumers aged 18 to 34 use AI chatbots as their primary product research tool, ahead of Google Shopping and Amazon search combined for that cohort. Adobe Analytics measured a 693 percent year-over-year increase in traffic to retail sites from generative AI tools during the 2025 holiday season, per the company’s January 2026 holiday shopping report.

The Infrastructure Backing the Rebuild

Amazon’s $200 billion capital expenditure plan for 2026 funds the AI infrastructure underlying all of the above. Jassy wrote in the letter that the investment is backed by committed customer demand, citing a partnership with OpenAI valued at over $100 billion as one disclosed example, with additional agreements described as in process or completed but unannounced. AWS’s AI revenue run rate reached over $15 billion in Q1 2026, which Amazon described as nearly 260 times the growth pace AWS recorded at a comparable stage in its development. Amazon added 3.9 gigawatts of power capacity in 2025 and plans to double total capacity by the end of 2027, with Jassy noting the segment is still supply-constrained despite that pace. The custom chip business, spanning Graviton, Trainium, and Nitro, reached an annual revenue run rate of over $20 billion growing at triple-digit percentages year over year.

Jefferies analysts, responding to the letter, described the resulting free cash flow pressure as a timing effect consistent with AWS’s historical investment pattern, projecting a potential inflection in both free cash flow and return on invested capital in the 2027 to 2028 period. Amazon’s free cash flow fell from $38 billion to $11 billion in 2025, driven by the $50.7 billion year-over-year increase in property and equipment purchases.

The AI Amazon is deploying inside its retail experience is the same stack being sold to enterprises at a $15 billion annual run rate, with Amazon as the largest customer of its own system. Jassy wrote in the letter that rebuilding from first principles is “what we’re doing in all of our consumer experiences,” and acknowledged it may take time to find experiences that outperform what currently exists. Rufus reached 300 million users and $10 billion in tracked annualized sales while that broader rebuild is still in progress, which makes it the most useful available signal of what the next version of Amazon’s store is being trained to reward.

Conversations On Retail

Conversations On Retail is a gathering place and resource center for retail and CPG executives, built to make it easier to stay current, discover the technologies and solutions shaping the industry, and connect with the people driving it forward.

We publish news, views, and reviews from staff editors, contributing experts, and trusted partners. Some articles are developed internally, while others are submitted by industry contributors or adapted from interviews and recorded conversations with industry leaders.