How Does Amazon Rufus Choose Products, and How Should Sellers Prepare?

What happened to Rufus?

Amazon launched Rufus in beta in February 2024, describing it as an expert shopping assistant trained on its product catalogue, customer reviews, community Q&As and information from across the web. Usage grew quickly: Amazon said more than 250 million customers used it in 2025, with monthly users up 140% year on year, and later reported about 300 million customers and nearly $12 billion in incremental annualised sales for the year.

On 13 May 2026, Amazon launched Alexa for Shopping in the US, bringing Rufus and Alexa+ together. There the standalone Rufus chatbot was discontinued, but Amazon said it would use Rufus's recommendation features and shopping history in Alexa for Shopping; other Amazon marketplaces still offer Rufus under its own name. The assistant appears in the search bar on Amazon's website and app, on Echo Show devices, and remembers conversations across them. Everything below applies to Alexa for Shopping as much as it applied to Rufus.

How Rufus, Alexa, and Sparky decide which products to surface

None of the three publishes a ranking formula, but their makers describe the inputs:

Two ideas explain most of what these assistants do. First, they answer needs, not keywords. Amazon's research on a system called COSMO describes mining customer behaviour to learn commonsense links between products and intentions, so a question about "shoes for standing all day at work" can be matched to cushioned, supportive shoes even if no listing uses those words. Second, they lean on evidence. Reviews are summarised into claims about fit, durability or ease of use, and listings whose own details contradict those claims lose out. Amazon's Alexa chief said in May 2026 that access to customer reviews and the catalogue is what gives its assistant an edge over rivals that scrape the web.

Does Amazon Rufus use customer questions and answers from listings?

Yes. Amazon's launch announcement listed community Q&As among Rufus's training sources. Since then, sellers have reported that product pages increasingly show AI-generated answers where the questions section used to be, and Amazon now points shoppers to its assistant for product questions. In practice, the answers shoppers see draw on your listing content, reviews and past Q&A. So the questions customers ask are best answered in the listing itself, where you control the wording, rather than left to the assistant to infer from reviews.

How to adapt Amazon product content for AI search

Write for the questions shoppers ask:

• Title: brand, product type and the attributes that define it, such as size, material or compatibility. Follow Amazon's title rules and avoid repeated keywords.
• Bullets: one benefit per bullet, stated as a fact with the use case, for example "Holds a 16-inch laptop and fits under airline seats".
• Attributes and backend fields: fill every structured field for the category, including target audience, special features, materials and dimensions. These are the facts assistants use to filter.
• Description: answer the objections buyers raise in reviews, such as noise, sizing or assembly time.
• Images: label key facts on infographics, and make sure the same facts appear in text, since text is what assistants read most reliably.
• Reviews: ask every buyer for an honest review, for example with Amazon's Request a Review button, offer nothing in return, and respond to recurring complaints by fixing the product or the listing.

How should Amazon sellers prepare for Rufus AI?

Treat it as preparing for Alexa for Shopping, which inherits Rufus's role:

1. Collect the questions. Read your reviews, past Q&A and returns reasons, and list the questions shoppers ask before buying.
2. Answer each one in the listing, in bullets, description or A+ Content.
3. Test the assistant. Ask Alexa for Shopping (Rufus outside the US) the questions a buyer would ask, such as "best carry-on backpack for a 16-inch laptop", and see whether your product appears and how it is described.
4. Correct contradictions. If the assistant repeats a claim from reviews that your listing does not address, such as "runs small", add sizing guidance.
5. Keep ads separate in your planning. Amazon Ads sells sponsored prompts that appear in shopping results and on product pages and open a conversation with the assistant. They buy attention; they do not fix a listing that fails to answer questions.
6. Keep price and stock consistent, since assistants avoid recommending products they cannot confirm are available.

A+ Content versus Brand Store pages for Amazon Rufus visibility

A+ Content sits on the product detail page, so it is part of the information attached to each product. Amazon has not documented exactly how its assistant weighs A+ Content, so put the essential facts in the title, bullets and attributes as well, and use A+ modules to add comparison charts, use cases and answers to common questions in readable text, with descriptive image text where the module allows it.

A Brand Store is a separate set of pages for browsing your range. It helps shoppers who are already looking for your brand, and it gives you a place to explain collections, but it is not where product-level answers come from. If you have limited time, improve product detail pages first, then A+ Content, then the Brand Store.

How can a Walmart Marketplace seller improve citations in Walmart Sparky answers?

Walmart describes Sparky as an assistant that helps shoppers find items and synthesises reviews, and it has made its catalogue available through partnerships with other AI platforms. Walmart's own Listing Quality score gives the clearest guide to what it values: content, discoverability, offer, and ratings and reviews. To improve your chances:

• Raise Listing Quality: complete every attribute, add strong images and rich media, and write specific titles and descriptions.
• Compete on offer: price, shipping speed and availability are part of the score; fast, reliable fulfilment helps.
• Build reviews through Walmart's own channels: Walmart asks buyers for reviews itself, and sellers may use its Review Accelerator programme or approved review syndication, never unsolicited buyer contact, package inserts or anything offered in return.
• Answer questions in the listing, as on Amazon, so Sparky can match your product to needs such as "best laptop for an art student".

What should sellers measure?

Visibility inside marketplace assistants is harder to track than web AI answers, because most visibility tools focus on ChatGPT, Google and Perplexity. Some specialist tools track Amazon's and Walmart's assistants, as covered in Peec AI vs Goodie for ecommerce teams. Otherwise, run a monthly set of buyer questions in each assistant, record which products appear and how yours is described, and compare with sales and conversion data. The wider ecommerce picture is in how ecommerce businesses should approach GEO.

What might this look like in practice?

Consider an illustrative travel bag brand, Vaskfen, selling a carry-on backpack on Amazon and Walmart. When asked for a "carry-on backpack that fits a 16-inch laptop", Alexa for Shopping named two rivals and described Vaskfen's bag, from reviews, as "tight for larger laptops". Vaskfen's listing never gave laptop compartment dimensions. The brand added exact compartment sizes to the title, a bullet and the attributes, updated an infographic with the same numbers in text, and added a comparison chart to its A+ Content. On Walmart, it filled missing attributes and raised its Listing Quality score. Two months later, the bag appeared in both assistants' answers to laptop-size questions.

What do people ask about Rufus and marketplace AI assistants?

Is Rufus still available?

In the US, Amazon replaced the standalone Rufus chatbot with Alexa for Shopping in May 2026, which uses Rufus's recommendation features. In other marketplaces, Rufus is still available under its own name.

Does Alexa for Shopping use reviews?

Yes. Amazon says access to customer reviews and its catalogue is central to the assistant.

Can I pay to appear in Rufus or Alexa answers?

Amazon Ads offers sponsored prompts that open conversations with the assistant, but organic answers depend on listing content and reviews.

Does Walmart Sparky work like Rufus?

It plays a similar role, using Walmart's catalogue and reviews to answer shopping questions.