Amazon puts an AI shopping buddy in your pocket

Searching on Amazon has always required you to know exactly what you are looking for. If you just have a rough idea, like setting up a backyard garden, you usually have to open ten tabs just to figure out what supplies to buy.

Amazon just launched an AI-powered shopping assistant named Rufus inside its mobile app. Starting today, a subset of U.S. customers can chat directly with the bot to find products, compare specs, and get personalized recommendations before buying.

A shopping chatbot trained on customer reviews

Amazon told TechCrunch that Rufus is powered by a new internal large language model built specifically for commerce. It was trained on Amazon's massive product catalog, customer reviews, community Q&As, and broader web data.

To use it, shoppers tap a new icon in the app navigation bar. A chat window slides up from the bottom of the screen, where you can type or talk.

Instead of searching for exact product names, you can throw general problems at it like "What do I need for cold weather golf?" or "What are good gifts for Valentine's Day?"

You can also get granular while looking at a specific item, asking things like "What is the difference between trail and road running shoes?" or whether a specific pair of boots is durable. When you are done, swiping down dismisses the chat back to standard browsing.

Amazon confirmed Rufus will launch without ads, though it noted more elements will be added over time. Customers can rate answers with thumbs up or down.

Why it matters

We think this is a major shift in how people will interact with online retail. Traditional e-commerce relies on users guessing the right search keywords and wading through sponsored listings. A conversational assistant flips that, turning vague lifestyle needs directly into a curated cart.

For everyday shoppers, it looks genuinely useful for product categories you know nothing about. Instead of reading dozens of fake or rambling reviews, an AI summarizing real sentiment saves real time. If you have access to the beta, it is well worth trying on your next purchase.

The real test will be trust. Amazon admitted generative AI will make mistakes, and its previous enterprise chatbot Q suffered from hallucinations and leaked internal data. If Rufus gives bad recommendations, users will quickly ignore it.

There is also the business model question. Amazon's ad business brings in tens of billions of dollars. If shoppers stop browsing search result pages and just buy whatever Rufus suggests, Amazon will eventually face intense pressure to monetize Rufus, which could quickly destroy the assistant's objectivity.