AI shopping assistants triple traffic, hitting retailers

Traffic from personal AI shopping assistants has more than tripled across Forter's network in a month, following the arrival of tools including Muse and Instinct, the commerce platform said.

Retailers are now dealing with a sharp rise in transactions initiated by so-called agentic assistants as they approach the holiday shopping period. Forter tracks payments, identity and fraud prevention across a network linking more than 2 billion shoppers with nearly 1 million businesses, covering half a trillion dollars in annual retail transactions.

Its data showed agentic order volumes rising steadily through late summer. From August to September, orders increased 24% month on month, or 48% including Instinct and Grok Bot after Forter detected their first orders at the start of the month.

Muse, launched later in the month, had a bigger immediate effect. Forter began detecting Muse orders on September 21, and agentic orders rose 229% month on month that week. Although that initial lead later eased, Muse still accounted for 46% of agentic orders in the final week of the month.

The shift adds a new layer of complexity for online retailers. Unlike a conventional large language model that responds to prompts with text, these assistants can store logins, two-factor authentication details and card information, then carry out parts of the buying process on a shopper's behalf through services such as WhatsApp or iMessage.

Retail pressure

That means merchants may face both a higher volume of automated shopping activity and new questions over whether those agents should be treated as trusted buyers at checkout. One issue, Forter said, is that some assistants will crawl a retailer's website if it has not been adapted for agentic traffic, slowing interactions and increasing the risk of errors.

"Agentic traffic has more than tripled in a month across Forter's network since Muse and Instinct launched.

"What makes Muse different from other agentic pilots is that these agents will default to crawling a merchant's website if it's not optimised for agentic traffic. Crawling the site takes significantly longer, is more costly and is very error prone. Merchants are effectively opted in by default, whether they've prepared for agentic traffic or not.

"This traffic surge creates two problems for merchants: ensuring their infrastructure is set up to support agentic transactions, and deciding whether they can verify and trust the agents attempting to check out.

"The backdrop to all of this, of course, is that these developments come right before the holiday shopping season. Agents present both a huge challenge and opportunity for retailers. Usually, merchants may have 6-12 months to react to changing consumer behaviour. Right now, the timeline is closer to 6-8 weeks," said Michael Reitblat, Chief Executive Officer of Forter.

The concentration of agentic shopping remains strongest in North America. Pilots tied to Google and OpenAI are limited to the US, while Muse is available only in the US and Canada, according to Forter. Among the services it tracks, ChatGPT Agent has the broadest geographic spread, with nearly a quarter of its orders coming from outside the US.

Buying patterns

Order values are clustering in a relatively narrow range. Purchases priced between $50 and $200 are the most common across all agents, according to Forter's data. ChatGPT accounts for the largest share of orders above $500, while Instinct and Grok Bot were recorded making the most expensive purchases overall.

Different agents also appear to be developing distinct retail niches. Forter said there is little overlap in shopping specialisms, with OpenAI leading in shipped retail categories including apparel, beauty, sporting goods and consumer electronics.

Consumer research cited by the company suggests willingness to use AI in shopping is moving faster than many merchants may have expected. In a survey of 2,000 UK adults on Christmas shopping habits, 53% said they have used or are open to using AI to find presents. The same survey found 22% believe AI would choose a better gift for them than their partner.

Even where shoppers are not ready to complete a full purchase through an assistant, retailers may still feel pressure to show they support the option. Forter said comfort with agentic shopping falls as consumers move deeper into the buying process, but 39% of respondents open to using AI also said they were more likely to shop with retailers that have AI purchasing agents.

For merchants, the issue is not only whether customers will hand over more of the shopping journey to software, but whether online storefronts, payment systems and fraud controls can distinguish between legitimate assistants and unwanted automated activity. Reitblat said the shift is happening on a compressed timetable, with retailers heading into their busiest trading stretch as agent-led transactions begin to move from pilot projects into everyday shopping behaviour.