Platform architecture is associated with the form and breadth of public engagement: algorithmic feed dependency in Chinese social media

Background Reliance on algorithmically curated feeds may relate to engagement with public issues beyond general use intensity, but whether this association varies across platforms and issue contexts remains unclear. Methods We analyzed 718 attention-check-valid responses from Chinese university students and recent graduates recruited online, including 508 users active on both Douyin and Xiaohongshu. Methods included OLS regression, paired comparisons with Holm correction, participant-clustered interaction models, repeated-measures measurement invariance tests, and item-exclusion sensitivity analyses. Results On both platforms, feed reliance was positively associated with educational issue engagement (Douyin: b = 0.136, p < 0.001; Xiaohongshu: b = 0.145, p = 0.001). Among four item pairs, only sharing/forwarding differed significantly after Holm correction (adjusted p = 0.009; liking/commenting: adjusted p = 0.062). Educational slopes did not differ significantly across platforms, though equivalence was inconclusive. Adjusted interaction tests showed steeper Xiaohongshu slopes for marriage/childbearing and social pressure issues. The educational-issue association persisted after excluding the continuous-scrolling item. Conclusion Findings reveal platform- and topic-specific correlations and offer preliminary evidence distinguishing perceived feed reliance from general use intensity. However, cross-sectional self-reports cannot establish platform architecture as the cause, and slope equivalence remains unestablished.