Amazon quietly kills Inspire feed and goes all-in on Rufus AI discovery

Amazon has shut down Inspire, its TikTok-style shopping feed, with no announcement. The swipe-and-buy experiment was Amazon's bet on social commerce, but the vertical video feed failed to drive the conversions the company expected: shoppers come to Amazon with purchase intent, not for passive scrolling.

The company is instead pivoting to AI-driven product discovery built around Rufus, its AI shopping assistant embedded in the Amazon app. Rufus factors in broader signals than keywords, including reviews, brand websites, and general internet presence, and proactively surfaces products using past purchases, browsing history, and price sensitivity. Early testing suggests personalization may override traditional ranking factors like review counts and keyword relevance.

For brands, the implications are significant: video reviews and influencer content will likely feed Rufus recommendations rather than live in a separate feed, well-branded products with strong off-Amazon content gain an edge, and sellers may need to optimize for AI recommendations rather than classic search rankings.

Rufus is also reading product images directly. It uses optical character recognition to pull specifications and instructions out of infographics and packaging photos, embeds product images inside its chat responses, and combines image content with extracted text to answer shopper questions. A tool from ProductPinion called ImageChek lets sellers preview how Rufus interprets their images.

The shift extends beyond Amazon. OpenAI's new Operator shopping agent foreshadows consumers delegating entire purchases to AI, which compares prices, checks stock, and orders based on preferences, potentially skipping Amazon's search and sponsored-ad system for routine goods. Analysts warn retail media could become obsolete for those transactions, with future 'digital shelf wars' fought over visibility inside AI ecosystems. Sellers are advised to invest in structured data, competitive pricing, brand loyalty programs, and AI-readable product feeds.

A product-research technique is also circulating: using Helium 10's Black Box module for brand analytics to surface keywords where the top three clicked products hold high click share but low conversion share, signaling unmet demand. In one bamboo-products example, filtering 2,183 keywords that way surfaced 215 actionable opportunities, including 'no tools bamboo shades,' which drew clicks but no conversions. Another seller used a database of 120,000 rising search terms to find three profitable products in a weekend.

In other headlines, the U.S. has ordered reciprocal tariffs on every country, top sellers are fuming about the growing share of Chinese sellers, and Amazon continues to struggle with physical stores. Shopify now holds 12 percent of the U.S. e-commerce market, with 875 million unique shoppers buying from its merchants.