Artificial intelligence has firmly made its way into helping merchants contest chargebacks, but even better use eludes many, finds a Chargeflow and Juniper Research report.

“The State of Chargeback Management: How AI Is Reshaping Performance” report found that merchants using AI to contest disputes win a median of 38%, compared with 27.75% not using AI.

That’s not the whole story, however. In both groups, those using AI and those not, win a median of 6.65% of the disputes they choose to fight, says Dan Moshkovich, Chargeflow vice president of marketing.

“AI is making merchants better at the fights they pick. It hasn’t yet changed which disputes they choose to fight, or the strategy and scale with which they fight them,” Moshkovich says in an email to Digital Transactions News. “That’s where most of the lost revenue sits unrecovered.”

Getting to that unrecovered revenue is important because 65% of respondents to the Chargeflow survey say the volume of chargebacks increased in the past 18 months and 90% of this group saw growth of 10% or more.

Moshkovich says the technology to help is ready, it’s the decisions about which disputes to prevent, refund, or fight and the manner in which they are fought that need addressing. A big part of that may come down to data.

“When asked, merchants rank data quality as the most important factor when choosing a provider and they’re right, because a decision is only as good as the dispute history behind it,” he says. “But looking at the latest in AI, we aren’t just looking at better data integrity in the representment delivery, but better data modeling that will deliver smarter representment itself.”

His advice is for merchants to start counting the chargebacks they choose not to contest, a stat many merchants may not have considered fully. “Confidence that the optimal representment is being delivered will improve merchant readiness to fight more disputes. Our Data Lab found ‘item never arrived’ disputes that went to a decision were won about 1.5 times as often with strong evidence, such as proof of delivery.”

Most automated representment services use templates. They pull the evidence they can find, drop it into a standard response and file it, he says. “That handles volume, but it doesn’t build the strongest case for each dispute,” Moshkovich says. “Merchants see the problem too. In Juniper’s survey, 45% of enterprise merchants said incorrect or unsupported evidence was a serious concern, and evidence gathering was the task they most wanted automated.”

Moshkovich cautions that different payment rails prioritize evidence based on their own criteria. “Knowing what evidence is most likely to move the needle on each rail delivers an outsized impact,” he says.

In other results, the report found that 39% of merchants would use a higher win rate and 34% would use more recovered revenue as reasons to justify spending on AI for their chargeback management.