Skip to main content
← Blog
4 min read

Processors Are Adding AI Fraud Screening Before the Holidays. Here's What Changes for You

By Yan Payments

What just changed at checkout?

Verifone turned on real-time AI fraud scoring, built on Visa’s Featurespace technology, across its payment gateways, and it’s part of an industry-wide shift toward AI-driven fraud screening ahead of the holidays. For most merchants this should mean better fraud detection without more friction at checkout, but only if the system is actually tuned for your business.

What did Verifone just announce?

In early November 2026, trade outlets including Finovate and The Paypers reported that Verifone deployed Featurespace, the fraud-detection technology Visa acquired in December 2024, across its payment gateways. The system screens in-store, online, and mobile transactions in real time using behavioral analytics that learn what a normal transaction looks like for a given merchant, card, or account, then flags anything that deviates from that pattern. Verifone hasn’t published its own detailed release on the rollout, so we couldn’t independently confirm specific performance numbers circulating in trade press, but the direction is clear: another large processor just added a real-time AI layer to its fraud screening.

Why does this matter if you don’t use Verifone?

Because it isn’t an isolated move. Visa bought Featurespace specifically to build this kind of scoring into its own network products and has already used it in A2A Protect, Visa’s fraud tool for account-to-account transfers. Mastercard has its own AI fraud push running in parallel. Most processors, large and small, have been layering machine-learning risk scoring into checkout for a few years now. If your processor isn’t Verifone, there’s a good chance it’s running something similar already or will be soon.

What’s actually driving the timing?

Fraud volume is climbing faster than fraud rates. Signifyd’s 2026 State of Fraud Report found fraud pressure across its network rose 33% year over year in the first four months of 2026, with card testing (bots trying stolen card numbers in small increments to find ones that work) up 175%. A separate Retail Dive report from May 2026 found attack rates actually fell 14% year over year, but chargebacks rose 56% over the same stretch, which suggests fraud is getting more concentrated and better at slipping past older rule-based filters. Processors are responding by swapping static rules, like blocking anything over a dollar amount from a new IP address, for models that weigh hundreds of signals per transaction and adjust as patterns shift.

Does more AI screening mean more declines, or fewer?

Both are possible, and which one you get depends on how the system is configured, not just whether it exists. A well-tuned model should catch more real fraud while approving more legitimate transactions that an old static rule would have blocked, because it can tell the difference between a new IP address alone and a new IP address paired with five other signals that actually indicate fraud. A poorly tuned one just adds another layer that blocks good customers.

Here’s why that matters in dollars: say you run 600 card transactions a week during the holidays at an average ticket of $45. The Merchant Risk Council’s 2026 report found roughly two-thirds of merchants see false decline rates of 2 to 10 percent of orders. Even at the low end, 2%, that’s about 12 legitimate transactions a week getting blocked, somewhere around $540 in sales that either get retried, go to a competitor, or just vanish. That gap is exactly what better-tuned AI screening is supposed to close, and exactly what poorly tuned screening makes worse.

What should you ask your processor right now?

  • Has anything changed in how transactions on my account are being scored in the last few months?
  • Can I see my actual false decline rate, not just my fraud rate?
  • When a legitimate customer gets flagged, what’s the fallback: a step-up verification, a manual review, or an automatic decline?
  • Who do I call if decline volume spikes unexpectedly during a busy week?

The bottom line

AI fraud screening is becoming the industry default, not an optional feature. That’s mostly good news, but it’s worth confirming your processor has it tuned for your business rather than assuming newer automatically means better. If you want someone to look at what’s actually configured on your account before the holiday rush hits, we’re happy to walk through it with you.

Sources: Finovate and The Paypers coverage of the Verifone/Featurespace deployment (reported around November 6, 2026); Signifyd 2026 State of Fraud Report; Retail Dive, May 2026 fraud attack rate and chargeback report; Visa newsroom, Featurespace acquisition and A2A Protect.


← Back to all posts