AI agent competition: how will bot competition change e-commerce?

On January 15, China's Alibaba announced that it has begun integrating its online shopping and travel services into its Qwen AI app to create a unified platform for its 100 million users. The new features are available for public testing in China. The company has also launched an invitation-only "task assistant" feature for more complex tasks such as booking trips or building web applications.

A month ago, Visa reported the first hundreds of transactions made by AI agents - the result of testing their use in commerce with more than 100 partners.

Visa predicts that by the 2026 holiday season, millions of consumers will be using AI agents to make purchases.

And McKinsey believes that by 2030, retail sales using AI bots worldwide could reach $3-5 trillion, by which time they could handle up to a quarter of the world's e-commerce sales, Deloitte writes, citing industry analysts.

What threats and opportunities do AI agents bring?

AI agents are intelligent personal assistants who know your habits and budget and can find the right product, compare prices and make the purchase themselves. They take care of the entire routine - from searching to filling out forms - with little or no input from you.

"AI agents will be a paradigm shift for e-commerce," says Nathan Feather, a researcher at Morgan Stanley Research. In the future, they will become "gatekeepers" in the world of commerce, according to McKinsey analysts.

The first such agents from AI leaders appeared last year. OpenAI in September launched Instant Checkout, an item purchase feature straight from ChatGPT. It's only available with a paid subscription for some items from Walmart, Shopify, Target and Etsy, CNBC writes. Perplexity unveiled Instant Buy in November in partnership with PayPal, and previously launched Buy With Pro for its paid subscribers.

Retailers and marketplaces are launching their own AI shopping agents, such as Amazon with its Rufus and Buy for Me, Walmart with Sparky, and eBay has its own assistant. Another major retailer, Target, has also launched an AI shopping assistant that suggests gift ideas based on user requests.

McKinsey experts wrote in October 2025 that it's still unclear what the customer journey will be in the future. But it is important that the mass distribution of AI interfaces ( ChatGPT alone has 800 million users per week, and Google's Gemini AI search reviews are used by at least 2 billion people per month) will lead to the fact that purchases, bookings, and so on will happen through AI channels.

A December Morgan Stanley survey in the US found that about half of users of AI platforms had used them to research and compare product prices in the past month, and 30-40% reported making purchases.

A study by OpenAI and Harvard economist David Deming, published in September 2025, found that about 2% of all ChatGPT requests are shopping-related.

Data from Similarweb, a web analytics company, shows that one in five referral link clicks to Walmart's website in August 2025 came from ChatGPT. Other retailers are also seeing growth, with data for fall 2025 showing ChatGPT driving more than 20% of referral traffic on Etsy, nearly 15% on Target, and 10% on eBay.

The speed at which ChatGPT has risen to become one of the top sources of referral traffic shows how AI is beginning to influence consumer behavior online, Digiday writes.

This is significant, especially when you consider that OpenAI has already begun testing ads in the free version of ChatGPT and in the cheapest paid plan, as reported by the Financial Times.

Despite the high rate of adoption, the AI agent market is still not massive. And according to a study by the University of Hamburg and the Frankfurt School of Finance and Management, traditional shopping methods still bring better conversion rates.

Researchers examined data from 973 marketplaces, analyzing more than 50,000 transactions made by ChatGPT conversions and comparing them to 164 million transactions from traditional channels.

ChatGPT is attracting more customers to websites, but most of them end up leaving without spending anything, Digiday points out. At the same time, the researchers recognize that although organic traffic from AI platforms performed worse than all channels except paid advertising on social networks, its conversion rate and revenue per session grew throughout the study.

Technology has other risks and limitations. They can make mistakes. Example: Scott Wingo, founder of e-commerce software startup ReFiBuy, was looking for a coffee machine through ChatGPT and was offered a Breville machine. When he clicked on the product, he saw an image of a garden rake.

This is a harmless error compared to the consequences of other potential "blunders" of AI agents: an incorrectly booked flight, overstocked goods, a purchase made without consent. When agents are interconnected across multiple systems, even minor errors can have a snowball effect, McKinsey writes.

It is not yet clear who will be financially liable in case of erroneous transactions - the developer of the AI agent, or the brand that implemented it, or perhaps the user? The EU AI law provides some clarity on high-risk systems, but its implementation is still evolving. In the US, fragmented regulation is creating a liability vacuum for companies, McKinsey writes.

There is another risk. When customers choose products, they rely on trust in a particular brand. But when an AI agent makes the purchase, trust becomes an abstract concept. The question arises - who do we trust if we don't make the decisions ourselves. The answer may be "no one," McKinsey believes.

Therefore, transparent policies in data development, storage and verification are needed, says Roger Roberts, partner at McKinsey.

Companies and marketplaces are now working with multiple protocols.

Anthropic has Model Context Protocol for AI agents to interact with external systems, Google has Agent-to-Agent, and OpenAI has Agentic Commerce Protocol, which allows bots to make purchases on behalf of a user using Stripe payment software.

On January 11, Shopify announced the launch of Universal Commerce Protocol, developed with Google to scale AI-enabled commerce. On the same day, Target also announced its implementation. Google says the protocol enables agent-based shopping directly in Google search and the Gemini app, and the platform is already supported by more than 20 companies, including American Express, Best Buy, Macy's, Mastercard, Stripe, The Home Depot, Visa, Zalando and others.

These are the first steps in licensing and verification. Those agents that use this protocol get access to catalogs and the ability to buy, those that don't - sellers will block.

Amazon has a different path. The company is actively developing its own AI shopping tools. Buy For Me allows users to shop on other retailers' websites, if Amazon doesn't sell the item directly, right in the Amazon Shopping app. Rufus is an AI assistant that recommends products and manages the shopping cart.

Rufus had more than 250 million customers last year, and monthly users grew 140% year-over-year, Amazon CEO Andy Jassy told analysts in October, according to Business Insider.

Amazon is expanding the presence of Rufus on its sites to encourage its use by users, CNBC writes. Without publicizing it, the company is experimenting with AI agents on subsidiary sites, for example, on the Woot sales site.

Amazon also recently added a feature that allows the chatbot to buy items once they reach a certain price, and has begun testing a feature that allows Rufus to create personalized shopping guides.

Last November, Amazon filed a lawsuit against Perplexity over the actions of shopping AI agents (CometAI). The lawsuit alleges that the startup secretly accessed the accounts of marketplace customers and passed off automated activity as the actions of a real person. As a result, Amazon cannot use its well-established model of personalized advertising.

Perplexity called Amazon's lawsuit an impediment to innovation and said user credentials are stored locally and never on its servers, Reuters paraphrased. Simplifying the shopping process means more transactions and satisfied customers, "but Amazon doesn't care, they're more interested in showing you ads," the AI developer said.

Amazon has a lot to lose. Revenues from its advertising segment totaled $17.7 billion in the third quarter of 2025, up 24% year-over-year. On total quarterly revenue of$180 billion, they account for nearly 10%. According to TD Cowen's forecast, Amazon's advertising revenue could almost double: by the end of 2025, it could reach approximately $68.2 billion, and in 2030 - $141.7 billion.

It turns out that the actions of other people's AI agents could be a hindrance to this growth.

Digiday wrote that due to Amazon's stance of refusing to introduce external shopping agents, Walmart was left without a major competitor in this new market.

Walmart doesn't have as much risk to its advertising revenue as Amazon. In its last reported quarter in November 2025, the company announced that revenue in its advertising segment jumped 53% year-over-year. But for the full fiscal year 2025, Walmart's advertising revenue generated $4.4 billion (just 0.65% of total revenue of $681 billion) - the bulk of which comes from the Walmart Connect platform. On it, merchants and brands buy sponsored spots to have their products show up higher in search and in ads.

Meanwhile, Walmart is also expanding its advertising business into the AI segment. In July, the company introduced Marty, an AI assistant that helps brands manage and optimize advertising campaigns, get analytics on their effectiveness and simplify the creation of advertising content.

"AI shoppers are threatening familiar advertising models, writes McKinsey.

Retailers that rely on advertising revenue on their platforms may face challenges as consumers move to chatbots, bypassing traditional advertising channels. This is understandable; an AI agent with the task of "finding the cheapest sneaker option" is unlikely to be distracted by banners with "unique" offers and fall for marketing tricks.

On the other hand, agent-based AI brings opportunities as well. Deloitte points out that in 2026, retailers will be able to use advanced AI tools to increase personalization and for audience analytics.

And McKinsey writes that companies could also charge a fee for analyzing data or providing premium services to consumers who use AI agents: creating sponsored "smart tips" and receiving commissions on transactions that users make with such bots.

This article was AI-translated and verified by a human editor