Of Agents and Humans: The New Reality of Holiday Shopping

November is the start of the holiday shopping season, especially for the millions celebrating Singles Day in China and parts of Southeast Asia, and Black Friday in the U.S. and some other parts of the world. 

While some will spend these holidays at physical, brick-and-mortar stores, with the attendant bleary-eyed early rising and long lines associated with them, even more will do their shopping online.

This year, however, digital commerce will take on a new twist as consumers will begin to deploy AI-powered agents to scour the internet, find the best deals and commence and finalize transactions on their behalf. Indeed, one may wonder whether bots will outnumber humans when it comes to online shopping this year.

A Growing Trend

While agentic commerce was a trickle last year, this year is shaping up to be much more significant. 

According to Salesforce’s 2026 holiday shopping predictions, 20% of all holiday ecommerce traffic is expected to originate from AI chat agents. That includes consumer-facing assistants, autonomous agents performing tasks and automated systems gathering pricing and product information.

Salesforce also found that half of shoppers have used an AI assistant at some point in their purchasing journey, while 74% say they trust product recommendations from AI chat.

Adobe’s research tells a similar story. In September 2026, AI-driven traffic to U.S. retail sites grew 71% compared with the previous year. More importantly for retailers, visitors arriving from AI sources converted into buyers at a rate 90% higher than other traffic sources.

Visa’s 2026 holiday shopping survey found that 54% of American adults have already used AI for at least one shopping task, compared with 47% last year.

The numbers suggest that AI is rapidly becoming part of the normal shopping experience. But there is an important distinction between using an AI assistant to research a product and allowing an autonomous agent to complete a purchase.

Someone asking ChatGPT to recommend a television is still largely making the purchasing decisions themselves. An agent that compares models, selects a retailer, enters payment information and places an order is doing something fundamentally different.

And that brings us back to the question of whether AI agents will actually outnumber human shoppers this holiday season.

Will There Be More Bots Than Humans?

Probably not, at least when it comes to completed purchases. There is currently no reliable industry-wide measurement showing that autonomous agents will complete more holiday transactions than human shoppers in 2026.

But when it comes to the number of interactions retailers receive, the answer becomes more interesting.

An individual consumer might visit five websites before purchasing a new laptop. An AI agent working on that person’s behalf could potentially examine dozens of retailers, compare hundreds of listings and repeatedly check prices and inventory before making a recommendation.

Multiply that activity across millions of shoppers, and the volume of automated interactions could become enormous, even if relatively few purchases are completed autonomously.

Retailers also face the challenge of distinguishing legitimate shopping agents from traditional bots. A pricing bot scraping thousands of product listings might behave similarly to an authorized shopping assistant looking for the best deal.

Both operate at machine speed. Both may access large amounts of product information. Both may interact with websites and APIs without a person directly controlling every action.

The difference is that one may be acting on behalf of a legitimate customer, while the other may be collecting competitive intelligence, abusing promotional offers or attempting fraud.

The Fraud Problem With Agentic Shopping

Consider a customer who authorizes an AI assistant to spend up to $200 on holiday gifts. The agent is allowed to compare products, select retailers and complete transactions within that budget.

Now imagine someone compromises the agent’s credentials or attempts to impersonate it. Could that attacker place orders using the customer’s stored payment information? Could they change shipping addresses, redeem loyalty points or purchase gift cards?

These are not entirely new forms of ecommerce fraud. Account takeover, stolen payment credentials and loyalty abuse have existed for years. What changes is the speed and scale at which automated systems can carry out those activities.

Traditional fraud controls often examine signals such as device information, location, transaction velocity and unusual purchasing behavior. Those signals remain valuable, but they become harder to interpret when legitimate customers are deliberately using software to shop at machine speed.

An agent making 50 product requests in a minute might be performing exactly the task its user requested. Another automated system exhibiting similar behavior might be probing for vulnerabilities or attempting unauthorized transactions.

Knowing Who, or What, Is Shopping

Retailers need to establish not only whether an account belongs to a legitimate customer, but whether the agent accessing that account has permission to perform a particular action.

At Microblink, we refer to this broader approach as Know Your Actor. The idea is to establish confidence in the identity of the person, bot or AI agent involved in an interaction, understand its authority and evaluate whether its actions are consistent with that authority.

A customer might authorize an agent to search for products and compare prices without granting permission to purchase anything. Another might allow purchases up to a specified amount but require confirmation before changing a delivery address or using a different payment method.

Retailers will also need to consider how these permissions are enforced across their systems. Shopping carts, payment processors, loyalty programs and customer accounts may operate through different platforms and APIs. If each system makes its own assumptions about what an agent is permitted to do, fraudsters may find opportunities to exploit the gaps.

The goal should be to let legitimate agents complete authorized tasks while introducing additional verification when an action falls outside the customer’s established permissions.

A Holiday Shopping Season Unlike Any Other

This holiday season probably won’t mark the moment AI agents complete more purchases than human beings. But it may be remembered as the year automated shopping became a mainstream part of ecommerce.

Consumers are becoming comfortable asking AI to find products, compare prices and make recommendations. The next step is trusting those systems to act on their behalf.

For retailers, that presents an opportunity to reach customers through entirely new channels. It also introduces a difficult question: How do you welcome legitimate automated shoppers without giving fraudulent bots the same access?

The retailers best positioned for this transition will be those that can distinguish legitimate agents from malicious automation, verify what those agents are authorized to do and respond appropriately when something looks wrong.