Testing alternative main product images — including AI-generated ones (e.g., ChatGPT-generated variants) — even on a listing that is long-established, high-review, and already backed by an in-house design team.
The testing method used is an immediate manual swap rather than Amazon's built-in formal A/B split-testing tool: replace the live main image directly and watch day-to-day CTR shift against the prior blended average, rather than waiting on a formal split test to conclude ("I'm too impatient. So, I switch it right away"). This trades statistical rigor for a much faster read.
Broader point: main-image performance decays and needs continual disruption/testing even on listings that seem to have already "solved" their image — past success is not a reason to stop testing.