Conventional Kansei Engineering (KE) models treat morphological components as equally weighted predictors of affective response, implicitly assuming that fixation volume indexes perceptual significance. This study proposes an attention-weighted KE framework integrating AOI-based eye-tracking evidence into morphological variable weighting prior to form–emotion modelling. Ten Ming-style chair samples were encoded into 30 binary morphological parameters; affective ratings were collected via semantic differential survey (N = 389) and element-level fixation distributions via eye-tracking (N = 30), from which coefficient-of-variation-derived weights were assigned. A critical dissociation emerged: the backrest dominated fixation (50.01%) yet received the lowest perceptual weight, while the head-rail and front arm support received the highest — demonstrating that fixation dominance and perceptual weighting are decoupled constructs. Attention-weighted models achieved 83.3% predictive consistency across five of six Kansei dimensions (p >.05). Findings offer a quantitative method for integrating visual attention into KE modelling, with implications for perception-informed design.