MW Retailers keep tinkering with their AI shopping assistants, in search of better service
By Bill Peters
Amazon says it is combining its shopping assistant Rufus with Alexa+ to launch Alexa for Shopping
Signage for Amazon's "Alexa" is displayed on stage during an Amazon Devices event in New York in 2025.
While seemingly every retail executive has talked up their efforts to embrace artificial intelligence over the past few years, actual help from AI assistants remains glitchy. This year, the AI race for retailers has been increasingly focused on trying to make the bots better.
Amazon (AMZN) on Wednesday became the latest company to embark on that path, announcing that it would fold together its Rufus AI shopping assistant with its Alexa+ platform into a new service called Alexa for Shopping - just two weeks after Amazon CEO Andy Jassy sang Rufus's praises to Wall Street analysts.
The company told MarketWatch there was nothing specific about Rufus that wasn't working, but combining the platforms would just be more helpful to shoppers. When asked, the Alexa for Shopping chatbot on Amazon's website stated that it was "previously known as Rufus."
Either way, the move by Amazon comes as retailers, travel apps and restaurants continue to fine-tune their AI shopping assistants - or go back to the drawing board altogether.
Those agents haven't always offered the most consistent service. Concerns persist about consumer privacy and trust on AI shopping assistants, as well as hallucinations, errors on things like product availability and sizing, and suspicions of retailers' bias toward some items over others. Others have said shoppers still use chatbots mainly to learn more about products, rather than make purchases.
Sky Canaves, a retail and e-commerce analyst at eMarketer, said that along with those concerns, the process of shopping with AI assistants is cumbersome, often steering customers to retailers' websites to finish purchases.
Retailers' websites are often still more straightforward and convenient for shopping, she said. ChatGPT, she noted, is a blank screen with a prompt. They key to AI shopping, she suggested, was staying ahead of the shoppers.
"One of the bigger challenges for AI shopping assistants and the chatbots is really trying to get to true personalization that's both predictive and meaningful," she said. "Even Amazon, which has so many years of our purchase histories, doesn't really proactively recommend products that we don't know we might want yet."
Amazon on Wednesday said Alexa for Shopping would be available for U.S. customers on Amazon's website, its app and its Echo Show devices. That new service, the company said, would make shopping more personalized for consumers as they weigh what they want to buy.
Amazon said Alexa for Shopping would bring together Rufus's product knowledge and consumer history data with the personalized detail gathered by Alexa+. Along with purchases, the new platform could help customers with things like dishwasher repairs, or recommend supplies needed for events.
During Amazon's earnings call late last month, Jassy said monthly active users of Rufus had jumped 115% year over year. He said the chatbot had improved significantly since last year, and that the company would keep improving it.
However, Canaves said eMarketer's own surveys showed that the share of consumers who reported using Rufus regularly, or even knowing what it was, was relatively small. She also noted that Alexa, having been around for roughly a decade, was a well-known name under Amazon.
Still, she noted Amazon's bigger AI shopping ambitions. The company last year launched Shop Direct, which uses AI to help consumers find products across the web. And she said combining the resources of Rufus and Alexa+, which can search the broader internet, would allow Amazon's new AI assistant to offer more detail and less awkwardness in its responses.
"Rufus was really limited to the Amazon product catalog for its training, so it couldn't go outside of Amazon to pull in more context or answer questions that weren't related to products being sold on Amazon, for the most part," she said.
Shares of Amazon climbed 1.6% to close Wednesday at $270.13, just below the record closing high of $274.99 reached on May 6.
Retailers like Walmart $(WMT)$ and Target (TGT) have their own AI shopping tools, and travel platforms like Airbnb $(ABNB)$ are also leaning more heavily on AI. But figuring out what works for consumers and businesses has, if nothing else, been a process.
OpenAI this year halted its Instant Checkout platform, which allowed direct purchases through ChatGPT, according to reports, amid technical hurdles. Instead, OpenAI has opted to develop apps in ChatGPT that would take consumers to retailers' websites, like Walmart's, to finish purchases.
Restaurants like Starbucks $(SBUX)$ and Little Caesars have launched apps on ChatGPT to help diners with orders. However, those apps don't actually let consumers place orders directly, and instead direct them to the chains' websites.
Even as Jassy talked up the potential for Amazon's AI assistant, he noted shortcomings elsewhere in AI shopping. He said that while he expected Amazon to work with third-party AI agents to improve customer service, the service those agents provide wasn't all that great yet.
"They're not often able to get the pricing right or the product information right," he said. "They don't have any personalization data or any shopping history. And so we do want to see that get better with third-party horizontal agents."
Brian Chesky, CEO of vacation-rental app Airbnb, said during the company's earnings call last week that AI would help offer more granular detail on consumers and help summarize reviews of rental locations.
But he said that chatbots - for both travel and e-commerce - provide too much language in their responses to questions, in a business where consumers rely heavily on photos. Users have to type everything into the chat prompt, which can be less convenient than, say, clicking a price-range bar, when seeking price comparisons.
Chatbots, he added, still aren't great for comparing rental locations, particularly when a city like Paris has tens of thousands of homes listed on Airbnb. The bots also weren't great at deciphering things like maps.
"I don't think anyone has figured out AI for travel or e-commerce yet," Chesky said.
-Bill Peters
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
May 13, 2026 18:03 ET (22:03 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments