COVID-19 reinfections are prevalent, necessitating an urgent understanding of their cumulative, long-term impact in highly vulnerable populations. Evidence suggests that reinfections are associated with worse health outcomes; causal artificial intelligence (AI) modeling may provide additional insights. Using longitudinal symptom, biometric, and exertion data from 1,167 people with Long COVID and/or myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) in the Visible app, we investigated how cumulative COVID-19 infections may influence symptom trajectories. We evaluated five causal discovery algorithms with Non-combinatorial Optimization via Trace Exponential and Augmented lagRangian for Structure learning (NOTEARS) nonlinear emerging as optimal, showing strong predictive performance and structural validity. To evaluate pathways, we estimated interventional effects using the fitted structural causal model. Within our structural causal framework, our analysis suggests a 1.5% increase in post-infection symptom severity density and 4.9% increase in acute resting heart rate (RHR) across infections one to six. We estimate that pre-infection functional capacity (FUNCAP) may reduce post-infection symptom severity and symptom frequency by 27% and 21%, respectively, while pre-infection crashes may increase symptom severity by 8%. Due to symptom tracking requirements, this cohort was subject to a healthy user survivor bias, potentially underestimating true effects. These results indicate that reinfection count may exert measurable effects on symptoms and heart metrics, and pre-infection FUNCAP and symptom crashes may contribute strongly to post-infection outcomes. Therefore, managing FUNCAP and preventing crashes are possible strategies for people with Long COVID and/or ME/CFS. This study supports the utility of causal AI on patient-generated data to identify plausible causal pathways.