Subscriber OnlyTechnology

Rise of AI shopping ‘agents’ set to transform online shopping

OpenAI, Perplexity and Google create AI-powered features, leading brands to rethink how they sell products online

Artificial intelligence's move into the world of online shopping will force brands to review how they market their products. Photograph: iStock
Artificial intelligence's move into the world of online shopping will force brands to review how they market their products. Photograph: iStock

The world’s leading artificial intelligence companies are betting that shopping will become a big application of AI “agents”, in a shift that is set to transform the multibillion-dollar ecommerce sector.

OpenAI, Perplexity, Google and Microsoft have in recent months introduced AI-powered features that allow users to search for products through chatbots, with autonomous agents able to complete orders on behalf of consumers.

The rise of AI-powered agents has prompted sellers and brands to rethink how they sell products online, in particular how their products are spotted by AI systems and recommended by chatbots.

Advertisers are employing techniques – such as creating longer URLs with keywords or securing a mention on websites considered to be more authoritative by bots – to appear more prominently in AI-generated results.

Start-ups including Profound, fashion-focused Refine and Algolia have also emerged, offering the ability to monitor brand presence in AI chatbot responses.

Profound co-founder James Cadwallader said consumer behaviour was reaching an “inflection point” where people may no longer visit ecommerce sites.

“AI [agents and chatbots] steal or hijack that consumer from the brand,” he said. “Eventually, the consumer will only interact with the ‘answer engine’, and agents will become the primary visitors for websites and the internet.”

At the same time, AI services are being increasingly used as a search tool. Almost 60 per cent of European Google searches no longer result in a click, according to data from search engine marketers Semrush. Instead users rely on the AI-generated text “overview” that helps to answer their query.

Analysts at Gartner anticipate that traditional search engine volume will fall 25 per cent by next year, owing to the rise of generative AI chatbots and agents.

AI companies are launching ecommerce services to utilise these developments. OpenAI released an updated version of its Operator shopping system, renamed Agent, which can complete tasks within a web browser, highlighting shopping as one of its core use cases.

For example, when a user asked the system to purchase ingredients for a roast dinner with a dessert, the system navigated to a supermarket website, added the necessary items to the shopping cart and then handed control back to the user to complete the purchase.

OpenAI also plans to take a cut from online product sales made directly through ChatGPT by introducing an integrated checkout feature that will allow users to complete transactions without leaving its platform.

Perplexity recently launched its Comet argentic browser, which completes tasks across different apps on the desktop, such as calendars, websites and emails.

Microsoft’s “Action” feature can also browse the web for shopping requests. Google’s new “AI mode” brings up different product options and the company recently released an AI product tracker, which alerts consumers when the price of a desired item has dropped to a certain level.

“Shopping has to be a deeply personalised experience that resonates with users,” said Lilian Rincon, vice-president of product for Google Shopping. She added that Google’s latest features were designed to save consumers “time and effort”, preventing them from “having 20 tabs open, trying to research different products, and it being complex and painful”.

AI chatbots or argentic systems primarily select products for inclusion in their results by picking one of the top results in traditional search engines. Google, however, uses a combination of advertising and search results, alongside personal data already stored on users, to provide more tailored recommendations.

This means brands can sometimes target them through so-called search engine optimisation (SEO) techniques. However, marketers are adopting new methods to improve the likelihood of brands appearing in AI-generated results.

Nikhil Lai, an analyst at Forrester, said this required a focus on specificity in product descriptions and improving technical details, such as ensuring a brand’s website loads within three seconds, which has become increasingly important as bots prioritise sites that load quickly.

Hannah Chelkowski, co-founder of Blank Ventures, a Refine investor, said retailers were seeing a rise of “semantic search” in AI chatbots, where users will look online in broader terms, for example, for clothes for “a wedding in the south of France” instead of specific fashion items. This means product catalogues need to be “reorganised” to include text descriptions that match this style of searching.

Recent research from the University of Applied Sciences Upper Austria suggests chatbots are exposed to and can be influenced by advertising on traditional websites, with a preference for text over images. This means simpler forms of clear-text advertising may be better for brands seeking to appear within chatbot results.

Beyond influencing recommendations, Dimi Albers, chief executive at media agency Dept, said brands needed to prepare for a world where transactions occur on chatbots, rather than their own websites.

“It will be very complex for most brands to move fast enough, not only to be visible but also to understand that where you sell your goods is not going to be either on your platform, Amazon or retail, but in all the models that are popping up,” he added. “The agents are talking to each other instead of people necessarily doing some of those transactions.”

However, Inrupt, a start-up cofounded by world wide web creator Tim Berners-Lee and John Bruce, aims to give control of personal data back to consumers, storing such information in a digital wallet that the individual can choose to grant access to agents and later remove.

Bruce cautioned that AI agents mark a move to a world where “shops and brands no longer count” and consumer choice could be limited as the system selects products for them, rather than showing all of the options.

“[Consumers] trade utility for privacy, freedom of operation and choice ... Surrender a pair of shoes today and who knows what you are going to give up tomorrow?” – Copyright The Financial Times Limited 2025