Electric vehicles (EVs) replace combustion engines’ natural acoustic cues with synthetic alerts via low-speed Acoustic Vehicle Alerting Systems (AVAS). However, the sonic dimensions chosen - such as frequency bands, amplitude modulation patterns, and spatialisation - may not map onto the perceptual heuristics that visually-disabled pedestrians have developed for judging vehicle approach and speed. In this paper we extend prior Likert-scale analysis by examining free-text feedback from 31 sighted (“no disability,” ND) and 21 visually-disabled (VD) participants about EVs. Employing a mixed-methods approach - non-response analysis, sentiment scoring, and dictionary-based thematic coding - we find that VD pedestrians are nearly twelve times more likely than ND to raise audibility concerns (65% vs. 13%, Fisher’s p = 0.0002) and are far more prone to recount surprise-by-silence incidents (95% vs. 48%, p = 0.0006), despite similar overall sentiment toward EV sounds across groups. These results indicate that AVAS effectiveness hinges not on simply producing sound, but on selecting the right acoustic features that match the perceptual heuristics pedestrians use to detect approaching vehicles.