Design Arena argues side-by-side votes beat satisfaction scores for tracking AI models

Design Arena says a satisfaction score from April may not mean the same thing in July, even for the same model.

The company argues side-by-side votes hold up better over time because every model answers the same prompt for the same user at the same moment, so all models face the same bar.

The claim is Design Arena's own methodological argument; the report does not include supporting data for it.