When Bad Actors Look Legit, Retailers Need More Than AI

Retailers engaged in online commerce with third-party sellers are generally well equipped to catch clear violations, such as a prohibited product, a known bad actor, or an obvious policy breach. Technology can typically catch these violations, often before a seller even begins transacting on a platform.

The problem, however, is that bad actors are becoming increasingly sophisticated by using new techniques and technology to circumvent the automated filters meant to stop them. As this kind of problematic behavior becomes more prevalent and insidious, online retailers are being forced to re-examine their enforcement systems and explore new approaches to keep pace with the ever-evolving risk landscape.

Bad Actors Adapt, and AI Helps Them

Bad actors are nothing if not persistent. Once caught, they rarely go away; rather, they use their rejection to better understand how a marketplace detects problematic activity, then they adapt.

Here are examples of how some sellers offering violative products learn to adapt:

• Some use misspellings or coded language to get around keyword filters. Rather than market counterfeit Apple AirPods, a seller might offer “@pple A1rp0dz.”
• Some make listings intentionally vague or break up identifying information. Amid the GLP-1 craze, we’ve seen many sellers marketing “weight loss pens.”
• Some hide offerings in photos. For example, a seller might market a washing machine repair kit but show in photos illegal weapons parts, which can look similar.
• Some have innocuous listings but market violative products off-platform. For example, a seemingly harmless supplement seller may be marketing illegal peptides or controlled substances on social media and redirecting customers to their supplement storefront in a “buy this, get that” scheme.

These techniques aren’t new, but they've become even more consequential in the age of artificial intelligence. According to INTERPOL, AI-enabled fraud can be 4.5 times more profitable than traditional schemes, in part because criminals can automate more of the work and operate at scale. This problem of scale is the biggest threat many online retailers face today: What was once a normal game of whack-a-mole has become a game the size of a football field.

The good news is that there are solutions, but online retailers must be willing to rethink how they're approaching risk management.

Solution 1: Use AI for scale and experts for edge cases.

AI can take on much of the first-pass work that would be difficult for retail risk teams to manage manually. It can compare large volumes of activity, connect related information, and organize evidence so reviewers can spend more time on cases that actually require investigation. This is essential when the scale of problematic activity is a primary concern.

The value of AI isn't that it makes every decision automatically, but that it can triage seller applications and give enforcement teams a clearer picture of what deserves attention. From there, trained experts can weigh the evidence against policy and regulatory context and decide whether the signals add up to meaningful risk. The combination of AI technology and human expertise is powerful.

Solution 2: Implement ongoing monitoring and verification.

Verification cannot always be treated as a one-time event. Bad actors who look good at onboarding may have learned to circumvent filters. Or sellers who started off operating in compliance may shift their business models over time to include violative products. We’ve seen this, for example, with sellers of legitimate wellness products starting to offer unapproved peptides.

TikTok Shop offers a great model for ongoing verification. It recently introduced a process requiring U.S. sellers who change their legal business entity to briefly pause sales, submit new ownership and tax documentation, and reapply for certain qualifications. A seller may have passed verification before, but a meaningful change can create a new risk question that warrants scrutiny.

Solution 3: Go beyond your platform.

A violation that appears isolated on one marketplace may be part of a larger operation. Bad actors rarely operate on one platform: In addition to selling on one or more marketplaces, they typically market on social media, communicate through private applications, and chat about their offerings on forums. Cross-platform intelligence can help retailers identify connections and determine whether questionable activity reflects a broader pattern. Identifying networks gives retailers more context to make sound decisions rather than evaluating each account or listing in isolation.

Building out networks can increase response time, too. With more data helping to connect the dots, retailers can more quickly and confidently explain why they acted, especially in gray-area cases where one obvious violation may not settle the question. A defensible decision should be grounded in the evidence available at the time and made clear why it warranted action. That record gives risk teams a stronger basis for responding to questions or seller disputes.

The combination of bad actors and AI is a problem that can’t be ignored, but it’s also a problem that can be addressed. As evasion techniques continue to evolve, the retailers that thrive are the ones willing to rethink their risk strategies to get ahead of the curve.

Dan Frechtling is senior vice president of product and strategy at LegitScript, where he drives product innovation, grows the company’s expertise in merchant intelligence, and strengthens partnerships with leading platforms, marketplaces, and payment companies.