The study of attached algae communities' assembly mechanisms is essential for understanding the restoration of river ecosystem structure and function. This study analyzed attached algae communities in three representative rivers flowing through Guangzhou—Chebei River (natural habitat), Liede River (highly canalized), and Shahe River (mixed habitat)—that are subject to a low water level operation strategy. The analysis covered four consecutive flood seasons from 2020 to 2023, focusing on the recovery and environmental drivers of these communities under near-natural restoration measures. The results showed that 193 algal species across 6 phyla, 53 families, and 90 genera were recorded, with Bacillariophyta being dominant (51.26%). No significant spatial differences in algal standing stocks were found, while significant annual differences occurred, indicating higher temporal than spatial heterogeneity. Chebei River had the highest total number of species (184), while Liede River showed the highest abundance (2.17×109 cells/m²), biomass (3.16×103 mg/m²), Shannon diversity index (3.14), and Evenness index (0.66). Annually, Liede River had the largest increases in species and Shannon diversity index (41.43% and 15.58%, respectively), and Shahe River had the highest abundance and biomass increase (92.62% and 96.53%, respectively), with Evenness index remaining stable at around 0.6. These findings highlighted the effectiveness of near-natural restoration for attached algae recovery. Niche breadth and overlap analysis indicated that the attached algal communities in the studied rivers maintained overall stable niche breadth, with niche overlap between most species being below the competition threshold (0.6). These results reflect high resource-use efficiency, low interspecific competition pressure, and a relatively stable state of the community. The standardized random rate analysis revealed that either deterministic or stochastic processes could dominate the restoration process, and their relative contributions were regulated by environmental heterogeneity. Redundancy analysis further identified that total nitrogen, total phosphorus, dissolved oxygen, water depth and flow velocity are key factors affecting the rivers attached algae communities. The findings provide empirical support and a scientific basis for the effective implementation of near-natural restoration in urban river channels.