AI is shaping holiday shopping, but are retailers ready for the risks?
This holiday season, shoppers are leaning on AI more than ever, forcing merchants to navigate a new wave of challenges and rising fraud risks.
By Eyal Elazar, VP, Market Intelligence at Riskified
If the past decade-plus was about putting the mall in your pocket, 2025 is the year your pocket might start shopping for you. This holiday season marks the first truly AI-powered wave of ecommerce, one where convenience for consumers meets a new layer of complexity and risk for merchants.
Nearly three in four shoppers (73%) use AI at some point in their holiday shopping journey, from basic product discovery to experimenting with “buy it for me” features, according to a Riskified survey. AI assistants, once the stuff of sci-fi and tech demos, now compare prices, summarize reviews, create shortlists of products, and increasingly provide personalized recommendations to millions of gift buyers.
But with every new leap in convenience comes some ambiguity. AI assistants don’t carry IDs, don’t explain their decisions, and don’t step up to the counter the way human customers do. They obscure the person behind the purchase, leaving merchants exposed to more chargebacks, higher rates of returns, buyer remorse, and even new forms of policy abuse.
They’re also prone to mistakes, since they don’t have access to all the data or consume it ideally, which can lead to mistakes and inconsistencies.
As consumers hand more discovery and decision-making power to algorithms, merchants face a deceptively simple question: Who, exactly, am I doing business with?
When AI Buys, Who’s Accountable
Consumers are sprinting toward AI assistants because they promise friction-free discovery and decision support.
Here’s the breakdown: thirty-five percent use AI for product ideas, 37% rely on them to summarize reviews, and 32% use them to compare prices. Looking ahead to peak season, 58% say they’re likely to use AI to help with holiday gift shopping, and 70% are at least somewhat comfortable with AI involvement. In these behaviors, shoppers remain in control, with AI supporting and assisting their choices rather than acting independently on their behalf.
But as automation accelerates, visibility recedes. Thirteen percent of consumers have already completed a fully AI-referred purchase – and that share is rising, meaning merchants face more mediation and less direct interaction with consumers
This is where frictionless shopping can create blind spots for merchants. With minimal insight into who or what initiated a purchase, merchants struggle to validate intent and identity in real time – and when an AI assistant makes or influences a purchase, who’s accountable if things go sideways?
Those are the scenarios that are already emerging:
- A legitimate customer disputes a charge after the fact. Their argument is that the AI bought the wrong item, misunderstood a prompt, or simply clicked “buy” too fast.
- Hijacked AI credentials initiate unauthorized purchases. Fraudsters also leverage automation, and a compromised agent can buy with unmatched efficiency.
Together, these scenarios create a widening accountability gap just as purchase volume peaks.
The Agentic Era Is Here, And Fraudsters Are Already Cashing In
GenAI-powered shopping is booming. In Q3 2025, the number of merchants receiving orders from GenAI channels tripled compared to the start of the year. Traffic from GenAI-powered shopping tools was 1.1–1.7× riskier than typical search traffic. According to our early data, traffic referred by large language models (LLMs) for a major ticketing seller was 2.3× riskier than Google search traffic, while an electronics merchant saw 1.8× higher risk rates.
With that growth comes new challenges. Fraud rings are already using AI to probe weaknesses in identity verification, account security, and payment flows. AI-powered shopping also opens the door to automated arbitrage: assistants can quickly strip inventory and resell it through fraudulent storefronts — sometimes recommended by other assistants. Left unchecked, these tactics can distort pricing, erode trust, and drain revenue before merchants even realize what happened.
Protocols like the Agentic Commerce Protocol (ACP) exist to define safe and secure interactions between AI assistants and merchants. In theory, they aim to maintain merchant control and strengthen fraud defenses. In practice, adoption is limited: many assistants don’t fully implement the protocols, and real-world behavior often falls short of the intended safeguards. That gap reinforces why merchants must remain vigilant.
The bigger problem? Neither side has full visibility. Buyers may not fully understand AI’s actions, and merchants often can’t see the human behind the assistant. In that blind spot, disputes, chargebacks, returns, and even new forms of policy abuse can multiply.
Riskified’s research shows a clear paradox: shoppers worry about payment security, privacy, and AI mistakes — yet 36% already trust AI to influence purchases, nearly matching the 38% who trust in-store associates. Consumers may be conflicted, but they’re leaning on AI. Meanwhile, fraudsters rely on it without hesitation to scale their operation
Defining the Future of Trusted AI Commerce
This holiday season is the first stress test of what the future of online shopping will look like. The winners will be the retailers who treat this as a strategic shift rather than a seasonal anomaly. AI assistants are here to stay, bringing convenience and complexity in equal measure — and that complexity makes executive education essential. Fraud and risk leaders must ensure senior executives understand both the opportunities and exposures: innovation without strong risk controls is an open vulnerability. The bottom line: merchants who proactively integrate risk controls, data transparency, and shared intelligence — and who stay vigilant about AI’s evolving behaviors — will be best positioned to thrive in an AI-driven commerce world.
