What can Rufus do?

(And why a Welsh Corgi is about to change everything you thought you knew about selling on Amazon)

Speed read: TL;DR & FAQs

The Straight Answer

Amazon Rufus started as a conversational shopping assistant that answered questions about products. In early 2026, he is something considerably more. He now tracks prices autonomously, purchases products on behalf of customers when a target price is hit, reorders previous purchases from a single spoken instruction, and even shops on websites outside Amazon entirely. Amazon has invested $50 billion in OpenAI, sued a competitor for sending its own AI agent to shop on Amazon, and launched a browser-based AI that can navigate the web and complete purchases independently. This is no longer a chatbot with opinions about cheese sandwiches. This is the early outline of a system that will reshape how products are discovered, evaluated, and bought. Here is what we have been reading so you do not have to.

---

We have been walking this dog since January

When we first started writing about Rufus at the beginning of 2026, he was an enthusiastic but slightly clumsy companion. We took him out, asked him questions about products, watched him sniff around listings and come back with answers that were sometimes brilliant and sometimes baffling.

He told one seller their stainless steel water bottle was made of plastic. He marked bitterness as a positive attribute for coffee. He confidently stated a product did not contain nuts when the ingredients list said otherwise. He was keen, he was trying, and he was not always right.

We documented all of this. We wrote about what Rufus is, how he thinks, what he reads from your listing, and how Amazon search is shifting from keywords to conversation. If you have been following along, you already know more about this particular Welsh Corgi than most sellers will learn this year.

But here is the thing. While we were writing those pieces, Rufus was not sitting still. Neither was Amazon. And what has happened in the first few months of 2026 is significant enough that it deserves its own piece.

This is a long read. It has a serious bite - a bit like Rufus versus the postman.

But if you sell on Amazon and you want to understand what is coming before it arrives, this is the one to stay with. We have done the homework. We have read the earnings calls, the partnership announcements, the court filings, and the technology releases that most sellers will never see. We are putting a stake in the ground, right here in April 2026, so we have a clear reference point when we come back to check how things have played out.

Grab a cup of tea. Rufus has been busy.

He has stopped just fetching and started doing the shopping

When we first introduced Rufus, he was an assistant. You asked him a question, he went and found an answer. Sometimes he brought back the right stick. Sometimes he brought back a sock. But the dynamic was always the same: you asked, he fetched.

That has changed.

In late 2025, Amazon rolled out what the industry calls agentic capabilities. In plain English, Rufus can now do things, not just say things.

Here is what that looks like in practice. A customer tells Rufus they want a particular Bluetooth speaker, but only if it drops below thirty pounds. Rufus notes the price. He watches it. When it hits twenty-eight pounds, Rufus buys it automatically. The customer does not need to check back, set a reminder, or visit the listing again. The dog did the shopping.

Did you know?

Amazon launched auto-buy capabilities for Rufus in November 2025. Customers can set target prices and Rufus will purchase autonomously when the price drops. He also shows 30-day and 90-day price history directly in the chat interface.

Source: AWS Machine Learning Blog / Amazon Q4 2025 Earnings

It goes further. Rufus now handles conversational reordering. A customer can say something like: "Reorder everything we used to make chocolate brownies last week." Rufus connects the dots between past purchases and current intent, works out which products were involved, and adds them to the basket. If something is out of stock, he suggests an alternative.

He also adds products to the cart automatically during conversations, so a customer can go from asking a question to completing a purchase without ever leaving the chat interface.

This is not a refinement of a search tool. This is a shopping agent. The distinction matters enormously for sellers, and we will come back to why.

He has started visiting the neighbours

Here is where things get genuinely surprising.

Rufus is no longer confined to Amazon's own marketplace. Through two features called Shop Direct and Buy for Me, Amazon's AI can now show customers products from websites that are not on Amazon at all - and in some cases, complete the purchase on the customer's behalf without them ever leaving the Amazon app.

Let that settle for a moment. A customer searches on Amazon. Rufus finds what they are looking for on a completely separate brand website. Amazon's AI navigates to that site, fills in the customer's details using their stored Amazon payment and delivery information, and completes the purchase. The confirmation comes from the brand, but the entire experience happened inside Amazon.

Some merchants have reported finding out their products were appearing on Amazon through this system without having opted in. They only noticed when Amazon-originated orders started arriving.

Amazon is no longer just a marketplace. It is positioning itself as the intermediary for all online commerce - whether the seller is on its platform or not.

In March 2026, Amazon opened this up further by allowing merchants to send product feeds directly into the Shop Direct system. The infrastructure is being built for a world where Amazon's AI can shop the entire internet on your behalf.

For Amazon sellers, this creates a strange new dynamic. You are no longer only competing with other products on Amazon. You are competing with every product Rufus can find, anywhere. Your listing does not just need to be better than the one next to it on the search results page. It needs to be clear enough, complete enough, and compelling enough that Rufus recommends you over something he found on a brand's own website.

The money behind the training programme

To understand where Rufus is heading, you need to follow the money. And the money in early 2026 is extraordinary.

On 27 February 2026, Amazon announced a $50 billion investment in OpenAI - the company behind ChatGPT. This was part of a broader $110 billion funding round that also included $30 billion from Nvidia and $30 billion from SoftBank. OpenAI is now valued at over $800 billion.

Did you know?

Amazon now has major investments in both OpenAI ($50 billion) and Anthropic ($8 billion+) - the two leading AI companies. It also develops its own Nova family of AI models. Rufus currently uses a combination of Anthropic's Claude and Amazon's own Nova models to power different capabilities. Source: Amazon Q4 2025 Earnings / SEC Filings / CNBC

What does Amazon get for $50 billion?

Several things that matter for sellers.

First, Amazon and OpenAI are jointly developing customised AI models specifically for Amazon's customer-facing applications. That means Rufus, Alexa+, and whatever comes next will have access to some of the most capable AI technology in existence, tailored specifically for shopping.

Second, the timing is revealing. Within days of this deal being announced, OpenAI quietly removed its own Instant Checkout feature from ChatGPT - the one that let people buy products directly inside the chatbot. The Shopify-OpenAI shopping partnership appears to be over. OpenAI is now redirecting shoppers to retailer websites instead, and building dedicated retail apps inside ChatGPT rather than competing with Amazon head-on.

Third, Amazon gets exclusive rights to distribute certain OpenAI enterprise services through AWS. This is not just about shopping. It is about Amazon becoming the infrastructure layer for AI-powered commerce everywhere.

The practical translation for sellers: the AI behind Rufus is about to get significantly more capable, significantly faster. Amazon is not experimenting with this. They are building the foundations for what they clearly believe is the future of how people buy things.

The courtroom and the fence

While Amazon has been investing billions in making Rufus smarter, it has also been fighting to make sure he is the only AI shopping agent allowed on its platform.

In November 2025, Amazon sued a company called Perplexity AI. Perplexity had built a browser called Comet with a built-in AI shopping agent that could browse Amazon, find products, and make purchases on behalf of customers. Amazon told them to stop. Perplexity published a blog post titled "Bullying is not innovation" and carried on.

Amazon then built a technical barrier to block Comet from accessing its site. Perplexity released a software update within 24 hours that circumvented it - by disguising its AI agent as a regular Google Chrome browser.

In March 2026, a federal judge in San Francisco sided with Amazon and blocked Perplexity's agent from accessing Amazon accounts. The ruling established something that will matter for years to come: a customer giving an AI agent permission to shop on their behalf is not the same as the platform authorising that agent to access its systems.

The legal precedent is now set. Amazon controls who gets to shop on its platform - including which AI agents are allowed through the door. For now, Rufus is the only dog in the yard.

Amazon has blocked dozens of third-party AI agents from its marketplace, including OpenAI's ChatGPT. The message is clear: if AI-assisted shopping is going to happen on Amazon, it happens through Amazon's own tools.

Meanwhile, other retailers are taking a different approach. Walmart and Target have been testing partnerships with third-party AI tools. Google is building its own shopping agent features. The landscape is fragmenting, but Amazon is building a walled garden - and they are defending it in court.

What does this mean for sellers?

It means Rufus is not going to be one of many AI shopping agents competing for attention on Amazon. He is going to be the only one. And that concentration of power in a single AI system makes it even more important that your listing communicates clearly with him.

Alexa, Rufus, and the phone in your pocket

Rufus is not operating alone. He is part of a growing pack.

Alexa+ - the upgraded, AI-powered version of Amazon's voice assistant - became available to all US customers in early 2026, free for Prime members. It is not the clunky, command-response Alexa of old. It is genuinely agentic. It books restaurants, calls rides, orders groceries, navigates the web independently, and remembers your preferences across every Amazon service you use.

Did you know?

Alexa+ customers have two to three times more conversations than with the old Alexa. Shopping via Alexa has tripled. Recipe requests are up fivefold. Amazon reports that the percentage of users opting out of Alexa+ is in the low single digits.

Source: TechCrunch / Amazon Devices & Services

Alexa+ learns from your purchase history on Amazon, your viewing habits on Prime Video, your reading on Kindle, your listening on Audible. When she recommends a product or answers a shopping question, she is drawing on everything she knows about you - not just your search query. She shares this context with Rufus.

Then there is Nova Act - Amazon's browser-based AI agent, launched in general availability in 2025 and now powering key features of Alexa+. Nova Act can navigate websites autonomously, fill out forms, complete purchases, and handle multi-step tasks without human intervention. It is the technology that makes Buy for Me possible.

And there are reports that Amazon is developing a new AI smartphone, codenamed Transformer internally, designed to integrate deeply with Alexa+ and the broader Amazon ecosystem.

The picture that emerges is this: Amazon is building a world where its AI is with you everywhere. At home through Alexa, on your phone through the app (and potentially a dedicated device), on your computer through Alexa.com, and inside every shopping interaction through Rufus. Your preferences, your history, your habits - all connected, all informing what Rufus recommends and what Alexa suggests.

For sellers, this changes the context of every product recommendation. Your listing is not just competing for attention in a search result. It is being evaluated against everything Amazon's AI knows about the individual customer standing in front of it - their budget patterns, their brand preferences, their purchase history across every Amazon service.

What this means for your listing

Here is the thing. Everything we have just described - the agentic capabilities, the cross-platform shopping, the $50 billion investment, the legal battles, the ecosystem integration - all of it comes back to the same place. Your listing.

Because Rufus still has to read something. No matter how sophisticated the AI becomes, no matter how many billions Amazon pours into the technology, the system still needs clear, honest, complete information about your product to work with. That has not changed. If anything, it has become more important.

What usually trips sellers up at this point is thinking they need to do something radically different. You do not. The fundamentals we have been writing about since January still apply:

Write in natural language that answers real questions. Make sure your images show the product in context. Ensure your A+ Content contains genuine information, not just decoration. Keep your Q&A section populated with clear, helpful answers. Be specific about who your product is for and what situations it fits.

Amazon listings rarely fail because one element is wrong. They stall when the words, the visuals, and the search signals stop reinforcing each other.

But the stakes of getting this right are rising. When Rufus was just answering questions, a gap in your listing meant a customer might get an unhelpful answer and move on. Now that Rufus is actively purchasing products on behalf of customers, a gap in your listing might mean he never considers you at all. The customer does not scroll past your listing - they never see it, because the AI made the decision before they had the chance.

When Rufus is setting price alerts and auto-buying, your pricing strategy matters in ways it did not before. Customers are not checking back manually. They are telling the AI to watch for them. If your pricing is inconsistent or your margins are built on artificial markdowns, Rufus's 90-day price history will make that visible.

When Rufus is shopping across the entire internet through Buy for Me, your listing is not just competing with the other products on Amazon. It is competing with the version of your product that appears on your own website, or on a competitor's website, or on any store Rufus can find. If someone else describes the same product more clearly, Rufus might send the customer there instead.

This is where the Content Triangle framework becomes not just useful but essential. Content, Creative, and Visibility have always needed to work together. An AI agent that reads everything - your words, your images, your reviews, your A+ Content, your Q&A, your pricing, your competitor's listing, and potentially your own website - makes the cost of misalignment far more visible and far more immediate (need a reminder? Click here →).

And underneath all of this - the agentic buying, the cross-platform shopping, the ecosystem integration - sits something most sellers have never heard of. It is called COSMO. It is Amazon's knowledge graph, and it is the intelligence layer that allows Rufus to understand meaning rather than just match words.

COSMO is the reason Rufus can connect a customer asking about "something good for bad backs" with a product that mentions lumbar support, orthopaedic design, or ergonomic features - even when no one typed those words. It is, in many ways, the real brain behind the dog. We are going to give COSMO its own piece next, because understanding how it works changes the way you think about writing your listing. For now, just know it exists - and that it rewards sellers who describe the situations their product fits into, not just the features it has.

The honest baseline (because he still gets things wrong)

We would not be doing our job if we presented all of this without acknowledging something important. Rufus is still far from perfect.

Independent analysis suggests that AI shopping assistant recommendations match the actual best product only about a third of the time. Rufus hallucinates product specifications. He invents features. He pulls information from a single negative review and presents it as the definitive answer. He has been reported giving false disparaging information about products while praising competitors.

Sellers on Amazon's own forums describe him as "Always Idiotic" (their words, not ours). One particularly concerning report describes Rufus stating that a food product did not contain nuts when it clearly did - something that could have serious real-world consequences.

Amazon's own disclaimer still reads: "may not always get things right."

This matters because it would be easy to read everything above - the billions in investment, the agentic capabilities, the ecosystem integration - and assume

Rufus is already a polished, reliable system. He is not. He is a powerful, rapidly improving, occasionally brilliant, and still frequently wrong AI system that is being given more and more authority over purchasing decisions.

This is exactly why your listing content matters so much. A vague listing gives Rufus room to hallucinate. A clear, specific, well-structured listing gives him the information he needs to get it right. You cannot control the AI, but you can control what you feed it.

From house pet to Crufts contender

So where is this heading?

Right now, in April 2026, Rufus is a capable but imperfect shopping assistant with expanding powers and a technology stack that is being upgraded at extraordinary speed. He is the dog equivalent of a promising young competitor - good instincts, solid training, but still occasionally distracted by squirrels.

By this time next year, the picture will be different. The $50 billion OpenAI investment will have produced customised models specifically designed for Amazon's shopping experience. Nova Act will be more mature, more reliable, more capable. Alexa+ will have expanded beyond the US. The legal framework around agentic commerce will be clearer. The proportion of Amazon searches handled by conversational AI will have grown from 14% towards the projected 35% or higher.

Did you know?

Analysts at Evercore ISI project that Rufus and related AI commerce features will increase Amazon's retail gross merchandise volume by 4.44% by 2028 and enhance advertising revenue by approximately $4 billion over the same period. Amazon has stated plans to spend $200 billion in capital expenditure in 2026, with the bulk going to AI-related infrastructure.

Source: Evercore ISI / Amazon Q4 2025 Earnings

What to do with all of this

If you have read this far, you already know more about where Amazon's AI is heading than the vast majority of sellers on the platform. That is not a small thing.

The practical advice we have given in the previous pieces in this series still stands.

Run the Rufus test on your listing (you might like to revist that here →). Rewrite your bullets as answers to real questions. Fill your Q&A section. Make your A+ Content informational, not decorative. Be specific about who your product is for.

But add this to the list: pay attention. Read what Amazon announces. Notice when new features appear. Ask Rufus new questions every month and see if the answers change. The sellers who thrive in this environment will not be the ones with the biggest budgets or the most aggressive advertising. They will be the ones who understood what was happening early enough to adapt thoughtfully.

You are not behind. If anything, you are ahead - because you are reading this.

And if you are looking at all of this thinking it is a lot to navigate alone, that is what we are here for. A Discovery Call is a straightforward way to look at what Rufus is actually seeing when he reads your listing, and what you can do about it.

No pressure, no pitch. Just clarity.

The dog is getting smarter. The question is not whether to pay attention. It is whether you do it now, while there is still time to get ahead - or later, when everyone else has caught up.

We will be back. Rufus is not finished yet. And neither are we.

This is part of a series exploring what Amazon's AI shopping assistant means for your listings - and what to do about it. Previously: What is Amazon Rufus? and How does Amazon search work? Next: what Rufus reads from your listing and how to structure your content so he can actually use it.

TL;DR

Rufus is no longer just answering questions - he is buying products autonomously, shopping across the entire internet, and operating inside an AI ecosystem backed by $50 billion in new investment. Amazon has sued to keep competing AI agents off its platform, meaning Rufus is the only shopping assistant that matters on Amazon. His accuracy is improving but still imperfect, which makes clear, complete listing content more important than ever. This piece marks our April 2026 reference point. We will be back to check what changed.

---

FAQs

What are Rufus's agentic capabilities? Rufus can now autonomously purchase products when they hit a price target set by the customer, reorder previous purchases from conversational instructions, add items to a customer's basket during chat, track 30-day and 90-day price history, and shop on websites outside Amazon through the Buy for Me feature. These capabilities were introduced in late 2025 and early 2026.

What is Amazon's Buy for Me feature? Buy for Me allows Amazon's AI to complete purchases on third-party brand websites on behalf of customers, using their stored Amazon payment and delivery details. The customer never leaves the Amazon app. The brand handles fulfilment, returns, and customer service. Amazon opened this to merchant product feeds in March 2026.

Why did Amazon invest $50 billion in OpenAI? The investment, announced in February 2026, gives Amazon access to jointly developed customised AI models for its customer-facing applications including Rufus and Alexa+. OpenAI will use Amazon's Trainium AI chips and AWS infrastructure. The deal is part of a broader $110 billion funding round and signals Amazon's commitment to AI-powered commerce at scale.

Can other AI shopping agents buy things on Amazon? Currently, no. Amazon has blocked dozens of third-party AI agents from accessing its marketplace, including ChatGPT. In March 2026, a federal court ruled that user permission to an AI agent does not equal platform authorisation, blocking Perplexity's Comet shopping agent from accessing Amazon. Rufus is currently the only AI shopping assistant permitted to operate on Amazon.

Is Rufus accurate? Rufus is improving but still makes significant errors. Independent analysis suggests AI shopping recommendations match the best product only about a third of the time. Sellers report hallucinated specifications, incorrect feature attributions, and misleading information. Clear, specific, well-structured listing content reduces the likelihood of Rufus getting your product wrong.

How does Alexa+ connect to Rufus? Alexa+ is Amazon's upgraded AI assistant, now available to all US customers and free for Prime members. It shares contextual information with Rufus - including purchase history, viewing habits, and preferences across Amazon services like Prime Video, Kindle, and Audible. When Alexa+ recommends a product or Rufus evaluates a listing, they draw on this combined understanding of the individual customer.

Amazon Source References

• Amazon Rufus announcement: https://www.aboutamazon.com/news/retail/amazon-rufus

  
  

• Amazon UK Rufus launch: https://www.aboutamazon.co.uk/news/retail/amazon-rufus-launch-uk-generative-ai-shopping-assistant

  
  

• How Rufus scales with Amazon Bedrock: https://aws.amazon.com/blogs/machine-learning/how-rufus-scales-conversational-shopping-experiences-to-millions-of-amazon-customers-with-amazon-bedrock/

  
  

• Amazon Buy for Me announcement: https://www.aboutamazon.com/news/retail/amazon-shopping-app-buy-for-me-brands

  
  

• Introducing Alexa+: https://www.aboutamazon.com/news/devices/new-alexa-generative-artificial-intelligence

  
  

• Alexa.com launch: https://www.aboutamazon.com/news/devices/alexa-plus-web-ai-assistant

  
  

• Amazon Nova Act: https://labs.amazon.science/blog/nova-act

  
  

• OpenAI and Amazon partnership: https://openai.com/index/amazon-partnership/

  
  

---

About the writer

Hello - I’m Cindy, founder of Mrs Prime.

I started out as an Amazon seller myself early during covid (and still am going strong), which means I’ve experienced more than my fair share of the same frustrations most sellers run into at some point: listings that should work but don’t, tweaks that change nothing, and the occasional moment of wondering what Amazon is actually doing.

Over time I realised most listing problems aren’t caused by one obvious mistake. They usually happen because the different parts of a listing stop working together.

Through Mrs Prime I help sellers understand those patterns and fix the right things properly.

Read more about my journey and experience here →

If you’d like to be notified when new posts appear on Mrs Prime Unpacked, you can subscribe below.

Subscribe to Mrs Prime Unpacked

---