Volume of AI-generated content in disputes has roughly doubled in a year. Issuers still review the cases by hand, and they have learned to dismiss AI letters faster than the templates they replaced.
The chargeback system has been infiltrated by AI on both sides, and the volume of AI-generated content moving through it has roughly doubled in the past year. Banking apps now suggest dispute language to cardholders; consumer-facing tools draft polished fraud-claim narratives; processor dashboards auto-draft rebuttal letters, and a small industry of AI-rebuttal vendors sells representment text to merchants directly. The infrastructure is impressive, but the win rate is not.
Issuers still manually review the cases. The analyst on the issuer side reads dispute responses by hand, working from a structured questionnaire, with about 90 seconds per file. They have learned to recognise AI-generated letters on sight — same opening sentence patterns, same generic specifics, same hollow certainty about evidence that turns out not to be attached. AI-drafted rebuttals therefore lose at rates lower than the templated letters they were meant to improve on, because the analyst has been trained, through repetition, to discount them faster.
The defence against AI-drafted consumer disputes is not AI of your own. The merchant who wins is the one with better order-time records and a letter written by a human who has read the case. AI is useful in the data layer beneath the letter (gathering evidence, surfacing recent issuer decisions, formatting artifacts), but the artifact that goes to the analyst has to read like a person wrote it, because the analyst is a person, and people recognise machine-generated argument on first read. The merchants who realise this first will stop spending on AI-rebuttal services and put the money into the records that argue cases on their merits.