AI Fraud Prevention DataCan AI Fraud Detection Help Merchants Stop Post-Transaction Fraud?

Merchants are Not Using AI to Fight Fraud, Leaving Them Exposed to Chargeback Abuse

As generative AI becomes increasingly user-friendly, affordable, and accessible, it’s also making its way into the hands of bad actors who use AI to carry out sophisticated attacks.

In the past, companies could root out third-party fraudsters with relative ease by establishing filters at checkout that checked for standard anomalies; unusual order volumes, billing and shipping address mismatches, anomalous login patterns, etc. These indicators are becoming less reliable as fraudsters adopt increasingly sophisticated tools, though.

Fraudsters are leveraging AI to devise and launch more difficult-to-detect schemes. It’s not just humans that can be fooled: AI-facilitated synthetic identity fraud and deepfake-enabled biometric spoofing attacks are so complex that they can easily defeat legacy security systems that rely on deterministic logic.

One report from KPMG Canada reveals that 81% of surveyed businesses have been victimized by at least one successful AI-powered attack. Those targeted say that AI-generated email phishing campaigns (60%), deepfake document fraud (39%), and voice cloning attacks (24%) are the most common fraud vectors.

This presents us with the fundamental question: if fraud is getting smarter, are merchant defenses keeping pace? And, what can merchants do to meet these new threats?

Where Merchants Actually Stand

A sizable minority of merchants are taking a proactive stance toward emerging AI-enabled threats. According to new Chargebacks911 survey data that we just conducted, 27% of respondents say they use AI to detect fraud. But, that also means that nearly three-quarters merchants are virtually unarmed against the AI-enabled threats facing them.

36% of merchants say they do not use AI-based fraud prevention tools and have no plans to do so. A further 37% reveal that, while they plan to use AI to detect fraud down the line, they currently do not deploy the technology.

This is not all bad news; the 37% of merchants who say they are open to implementing AI fraud detection tools describe themselves as being only “temporarily” behind the curve. But, with the right support and choice of tools, these merchants can close the gap. So, the segment of respondents who are flatly unconvinced of the need to integrate AI into their fraud detection strategy sits at just over one-third of merchants. This number is still alarmingly high, though.

Chargeback Representment Management

 Source: 2026 Merchant Survey, "Do you manage chargeback representments in-house or do you use third-party support?"

The problem is compounded by the fact that a substantial portion of merchants elect to manage fraud and chargebacks in-house.

Data published by Nuvei and PYMNTS shows that 56% of merchants elected not to outsource their fraud prevention operations to specialized third-party providers. This is despite the fact that merchants who did outsource their fraud prevention reported a much lower rate of failed payment (5% versus 13% of attempted transactions).

Chargebacks911 survey data further found that 69% of respondents manage chargeback responses entirely in-house.

The typical merchant’s area of expertise is limited to areas such as their products and industry. It does not extend to fraud or dispute management. Few operators will have the expertise required to evaluate outside threats, understand the nuances of the dispute lifecycle, or keep up-to-date with changes to rulesets governing fraud and chargeback practices. In fact, just 17.4% of SMBs say they are “very” informed about card network rules, with the remainder admitting that they are “not at all” or, at best, “somewhat” informed.

From these insights, we can draw the following conclusion: the AI adoption gap isn’t happening in isolation. Rather, it’s part of a pattern of underinvestment in fraud infrastructure that leaves merchants systematically exposed.

AI Skeptics May be Making a Mistake

Before going any further, we should frankly address the (reasonable) skepticism that many still have toward AI.

Survey data compiled in late 2025 regarding general attitudes toward the use of AI shows that the technology does not have a net-positive degree of trust in any industry. But, trust is lowest as it pertains to money; only 19% of Americans say they trust AI in matters of finance. 

It’s possible that merchants with legacy fraud detection solutions in place may have legitimate reasons for refusing to switch. Customer data protection, for example; the idea that cybercriminals could target AI systems to access consolidated datasets containing sensitive customer information has been cited as a concern. There is also reluctance to reskill employees, or skepticism about the cost-effectiveness of the technology relative to traditional fraud detection.

Sellers who believe their legacy security systems function well enough may be unwilling to fix what is not perceived as broken. The difficulty in distinguishing hype from reality, especially when AI capabilities are sometimes disingenuously sold as a cure-all to a nuanced problem, may also turn off some decision makers.

None of these concerns are unreasonable. The counterpoint, though, is that what is at stake may already outweigh all of these concerns.

Merchants who do not at least begin to consider fighting AI-based fraud threats with AI fraud detection tools risk being exposed to an ever-increasing number of new fraud attack vectors to which they are fundamentally unprepared to respond. In effect, bad actors are playing offense; this gives them a first-mover advantage. Merchants will soon find that the reactive tools they trusted up to this point are getting increasingly unreliable.

It could also be much easier to achieve a positive return on investment than most of these sellers anticipated. According to data from LexisNexis Risk Solutions, US eCommerce merchants lose a total of $5.13 for every dollar lost to fraud in 2026; a 37% increase compared to 2022. Calculating the cost of an AI-based fraud detection solution against the true cost of fraud, which considers direct fraud costs in tandem with hidden costs like false declines, customer churn, chargeback fees, and processor penalties, can help merchants paint a more accurate picture of how much money is really under threat from AI-enabled fraud attacks.

What AI-Based Fraud Detection Actually Does

Traditional rule-based fraud solutions follow an algorithmic “if this, then that” approach. For example, rules can be set to automatically reject checkout attempts conducted by IP addresses outside the US or orders above $100 from accounts less than 24 hours old. They effectively rely on heuristics: they approximate the boundaries of “normal” buyer behavior. Anything exceeding those pre-programmed thresholds is flagged as fraudulent.

Naturally, this results in false positives. Sign-in activity that appears to be anomalous, for example, can in fact be a regular buyer who forgot a password and so attempted multiple logins in a row. False negatives, where fraudulent behavior is misidentified as legitimate, are also a possibility, especially when AI can be used to help an unauthorized user hide their activity as normal.

AI-based fraud detection is more advanced. Instead of relying on pre-programmed rules, AI systems ingest large amounts of data to uncover more reliable fraud indicators and trends. As fraud tactics evolve and more data is collected, these systems will learn to identify new fraud patterns and continually improve over time.

What’s interesting is that these fraud patterns are often subtle and only detectable at scale, so they can be difficult for human fraud analysts or rules-based systems to uncover. For this same reason, AI-based fraud detection systems are better at imparting nuance. This presents a range of potential benefits, including:

Closing the Gap

The 27% of merchants who are actively using AI to detect and prevent AI-enabled fraud are on the right track. Given that new fraud tactics are emerging daily, merchants need systems that can continually ingest new information, adapt to new trends, and detect novel threats. AI-enabled fraud detection solutions fit that bill and can help sellers safeguard revenue even as attacks become more complicated and harder to root out.

The 37% of sellers who plan to implement an AI-based solution down the line will need to do some planning, and do it fast. Taking those initial steps of developing a timeline, creating a budget, and consulting with third-party experts and in-house functional leaders can help turn an “eventual” ambition into something more concrete and actionable.

Last but not least are the 36% of merchants who have no plans to integrate AI into their fraud prevention strategy. These businesses will want to reexamine their motivations for ruling it out and, if possible, reconsider their stance towards upgrading. To make things more palatable, merchants do not have to commit to overhauling their existing infrastructure. Modular AI-based tools, such as fraud scoring systems or liveness checks at sign-up, can complement existing toolkits without replacing entrenched legacy solutions.

This whitepaper examines the status of AI adoption as a fraud detection tool using data compiled for the 2026 Chargeback Field Report. Eager to get a bigger-picture grasp on the state of chargeback management in 2026? Download the full report today.