A corrected headline is a small wrinkle, but the underlying chargeback mess is not. Chargeflow and Juniper Research surveyed 121 senior executives at U.S. and U.K. businesses with at least $100 million in annual gross merchandise value, and the useful takeaway is painfully clear: merchants are putting AI on the label without rebuilding the machinery underneath.
That distinction matters. Eighty percent of respondents say they use AI in chargeback management, while only 11% say the work is mostly or fully automated. That is not a revolution. It is a manual process wearing a newer hat.
The Win-Rate Gain Is Real but Not Yet a Breakthrough
Merchants using AI reported a median 38% win rate on the chargebacks they dispute, versus 27.75% for non-users. A 10.25-percentage-point spread is nothing to sneer at. If a company is already filing sensible disputes, better evidence assembly and faster case preparation can absolutely put recovered revenue back where it belongs.
But this is where the report gets irritating in a useful way. Both AI users and non-users disputed and won the same median share of eligible chargebacks: 6.65%. In plain English, AI appears to help merchants win more of the fights they choose, but it has not persuaded them to fight more of the worthwhile battles. Somebody still has to decide whether a case is worth filing, and apparently that decision remains stuck in the old workflow.
That is the last-mile problem. Businesses can brag about an AI layer all day long; if their teams leave most viable recoveries untouched, the financial result stays capped. A 38% win rate on a tiny slice of eligible claims may look good in a board deck and still leave a pile of money on the floor.
Enterprise Buyers Are Not Buying Fancy Demos
The survey says 39% of merchants would justify spending on AI through a higher win rate, while 34% picked more revenue recovered. Time savings and lower costs drew fewer than one in five responses. Frankly, that sounds right. Finance leaders do not need another software vendor promising fewer clicks. They want proof that leakage shrinks.
There is a hard expectation gap, though. Sixty-nine percent said an AI tool would need to increase their win rate by at least 11 percentage points. The reported AI advantage of 10.25 points falls just short. Close only counts in horseshoes and, apparently, sales presentations. Juniper’s conclusion that current gains are incremental rather than sufficient is the sober read.
- 72% rated provider data quality as very important.
- 70% assigned that same importance to security and compliance.
- Only 31% put total cost in the very-important bucket.
Those figures tell vendors where the real knife fight will be. Cheap is not the winning pitch when disputed revenue, sensitive transaction data, and compliance exposure are on the line. The provider that can show reliable data, defensible decisions, and measurable recovery has a shot. Everybody else is selling a shiny wrench to a company with no mechanic.
Holiday Claims Put Delivery Evidence Under a Microscope
The seasonal details deserve attention, especially for e-commerce operators heading into the December rush. Seventy-seven percent of respondents said chargebacks rise over the holidays, although 58% expect lower chargeback losses in the 2026 season than in 2025. Optimism is fine; it is not a control system.
Chargeflow’s Data Lab examined roughly 3 million disputes over the past three peak seasons. During the 2025 Black Friday-to-Cyber Monday weekend, 34.7% of chargebacks claimed an order never arrived, compared with 19.9% in a normal week. Fraud claims moved the other direction, falling to 27.2% from 37.8%. The company says that pattern also appeared in 2023 and 2024.
That shift should make merchants focus less on generic fraud theater and more on fulfillment proof, carrier handoffs, customer communication, and the evidence trail after an order ships. A customer saying an item never arrived is not the same operational problem as a classic fraud allegation. Treating every chargeback as one big fraud bucket is lazy management.
What This Survey Can and Cannot Prove
Chargeflow has more than 20,000 merchants using its platform and says it protects more than $50 billion in transactions, so it has every incentive to argue that better automation is the next purchase. Fair enough. The survey was also commissioned around its market, and it relies on responses from 121 executives and managers collected in September 2026. That makes it informative, not gospel.
The correction itself updated the subheadline and the opening sentence of the original release, a reminder to read vendor research with both eyes open. Still, the numbers point in the same direction: chargebacks rose over the prior 18 months for 65% of respondents, and nearly half put their annual cost at 0.5% to 1% of revenue. Friendly fraud is a major concern, with 43% believing it accounts for more than a quarter of their chargebacks.
AI is not the finish line in chargeback management; better case selection is where the recovery math finally changes.
For merchants, the practical question is brutally simple: does the system identify more winnable claims and turn them into filed, well-supported disputes? If the answer is merely “our win rate improved on cases we already filed,” that is progress, but it is not the promised land.