Holiday Shopping Goes Agentic: What That Means for Fraud and Chargebacks

Highlights:

• Holiday Volume: Equifax credit data shows the holiday quarter (October - December) adds $40.4 billion in bankcard balances, giving fraudsters volume to hide in.

• A Missing Signal: Fraud tools that read mouse movements and typing speed have little to go on when an AI agent buys through code.

• January Fallout: Order mistakes and unfamiliar seller names can push real shoppers to dispute legitimate charges, which become chargebacks for merchants.

• Comprehensive Verification: Equifax payments fraud solutions evaluate device details, IP addresses, and location data in real time to verify valid buyers.

Holiday shoppers have a new helper this year: AI agents that compare prices, build carts, and in some cases complete the purchase. For merchants, that convenience arrives at the busiest and most exposed point on the retail calendar. As transactions move from human-driven screens to automated software routines, traditional fraud detection faces an unfamiliar challenge. Determining whether a digital purchase is valid now requires examining transaction data rather than looking for a person behind a keyboard.

According to Equifax credit data examining fourth-quarter seasonal borrowing trends, outstanding bankcard balances rose by $40.4 billion during the holiday quarter (October - December 2025), lifting total balances to $1.124 trillion across 591.6 million active accounts. Credit card utilization climbed to a seasonal high of 21.2% in December, while bankcard severe delinquencies (defined as accounts 30 or more days past due) reached 3.1% by February. That sharp increase in volume creates ideal cover for automated fraud. Low-cost AI tools let bad actors run card-testing scripts and launch fake storefronts in minutes, blending into legitimate seasonal traffic.

When the Shopper Has No Mouse to Move

Most fraud defenses were built around how people behave on a screen: how they move a mouse, tap a phone, or type. An AI shopping agent often skips the screen and connects to a merchant's systems through code. With no human signals to read, older checks can end up blocking legitimate customers or waving automated fraud through.

The scale of the problem is large. In 2025, automated programs accounted for more than half of web traffic, and harmful bots made up roughly 40% of visits (up from 37% in 2024).¹

4 Ways AI Shopping Can Go Wrong

Not every problem starts with a criminal. Shoppers who use AI tools to find deals and place orders can run into software errors, and the speed of those tools can trip a merchant's own defenses. Many of these problems surface as disputes once January statements arrive.

• The forgotten reorder: A shopper sets an agent to restock household items on a schedule, forgets, and tells the bank they never made the purchase.

• The wrong quantity: An agent misreads a listing and buys one item instead of a multipack. The shopper calls the bank rather than the store, and the merchant absorbs a chargeback instead of processing a return.

• The unfamiliar name: When one agent hands a task to another, an unfamiliar seller or service name can appear on the card statement.

• The promo-code rush: Agents can test promotional codes far faster than any person, which can trip store security and cancel valid orders.

Catching Fraud Without Slowing Real Shoppers

The answer is to judge each transaction on its own evidence rather than on screen behavior. Equifax Payments Fraud on Kount® 360 evaluates device details, IP addresses, and location data in real time against identity records across a global data network. Its dual-layer machine learning pairs two approaches. "Experience" recognizes known patterns, such as scrambled billing addresses and recycled card-testing details. "Intuition" scans the network for new threats as they form. This dual approach can help reduce false declines by up to 65% and support approval rates up to 95%.²

After the holidays, Equifax chargeback prevention and dispute processing tools help merchants resolve valid claims quickly, protecting revenue without adding steps at checkout.

As more of the shopping journey is delegated to software, the question for merchants shifts from "Is this a person?" to "Is this purchase legitimate?" Answering it at the transaction level, rather than at the screen, lets merchants welcome helpful agents and keep fraudsters out.

Learn more about Equifax identity verification and payments fraud protection.

Sources:

¹ Imperva - 2026 Bad Bot Report

² Equifax - Payments Fraud