The New Economics of Merchant Fraud
How bad actors continue to evolve, and what you can do.
Something has changed in the past two years. The sellers applying to your marketplace, the businesses requesting payment processing: they don't look like they used to.
The obvious fraudsters are still out there. Increasingly, the applications that should raise red flags look completely legitimate. Professional websites. Realistic product photos. Plausible business descriptions. Clean documentation. Underneath, some of these businesses do not exist.
The asymmetry has shifted
A synthetic business can be registered for less than $150. An AI-generated website with product pages, About sections, and customer testimonials can be spun up in an afternoon. Between May 2024 and April 2025, scam pages created via generative AI quadrupled, hitting over 38,000 new fraudulent pages per day.
Meanwhile, the teams reviewing merchant applications are still working the same way they did in 2019. Open the website. Click through products. Check the registration. Google the company. Make a judgment call. This worked when creating a fake business required real effort. It does not work now.
The incentive problem
Every platform faces the same tension. Acquisition matters. Friction means lost revenue. At the same time, onboarding a bad merchant creates real costs: chargebacks, regulatory exposure, reputational damage.
Manual review becomes a bottleneck. Backlogs build. Reviewers face pressure to move quickly, which means less time to catch subtle signals. The sophisticated fraudsters slip through.
What manual review misses
A reviewer can spot obvious problems. Merchants who are actively trying to evade detection have learned what reviewers look for. They create a professional-looking site with enough legitimate products to pass a quick scan. They bury prohibited items deeper in the catalog or hide them behind login walls.
Manual review also struggles with what is not visible: are prices realistic for the category? Are product images stock photos used on dozens of other sites? Is there evidence the business actually operates? These questions require investigation, not observation.
Continuous risk
Initial onboarding is only part of the problem. A merchant legitimate at approval can change. Most platforms cannot afford to continuously re-review their merchant base. A good merchant gone bad can operate for months before anyone notices.
What deep investigation looks like
Roe's Merchant Risk Agent treats merchant review as an investigation, not a checklist. Each review takes 2–8 minutes depending on complexity. The agent analyzes the website at depth, examines listings and pricing, checks whether images appear elsewhere, and detects content present in the HTML but invisible to casual browsers.
Beyond the merchant's own site, it searches public records, social media, and news to build a picture of whether the business is what it claims to be. UBOs are screened against sanctions and adverse media. The output maps to your decision framework (approve, reject, or request more documentation), with specific findings and evidence attached.
