High-quality urban blue-green spaces are essential to enhance the well-being of residents. Public preferences in urban landscapes are closely related to visual behavior and can meet human needs. By combining objective eye-movement data with subjective evaluation data, public preferences for blue-green typical spatial waterfront parks were analyzed based on visual behavior. Taking Wuhan, a megacity in China, as an example, four types of landscapes were identified: waterscape, plantscape, hardscape, and integrated landscape. Using visual clustering to identify eye movement indicators, we analyzed the influence of element types, proportions and combinations on public preferences based on visual behavior differences related to landscape elements. The results indicated that: (1) visual behaviors differ among elements and preferences, arbor commands overall attention, and pavilion is the element attended to first. (2) The positive correlation between water and fatigue recovery increased by 0.163 after combination with reflection, while the correlation between arbor and fatigue recovery changed from 0.353 (p < 0.01) to − 0.251 (p < 0.01) after combination with shrub and grass. (3) Plantscape and waterscape were rated as more natural, intimate, and restorative, but the most element-rich integrated landscape had the highest beauty ratings. Hardscape had significantly lower ratings but were generally considered to be fully utilized spaces. (4) The most significant visual behavior factor affecting evaluations was fixation duration, and the most critical landscape elements were arbor and pavilion. This study reveals differences in people's visual perceptions and preferences for waterfront parks through eye-movement experiments, and could lead to more judicious landscape element selections and park space compositions.