Religious service attendance is associated with better well-being, but observational associations do not establish causation. We analyse six annual waves of the New Zealand Attitudes and Values Study ($N = 46{,}377$) to estimate causal effects of monthly attendance on 24 well-being indicators using target trial emulation. Deterministic ‘make everyone attend’ contrasts fail positivity: only 2–3% of non-attenders initiate attendance per year. We therefore estimate supported stochastic interventions ($\delta = 5$) among baseline non-attenders ($N = 38{,}477$) using a sequentially doubly robust estimator with cross-validated machine learning. Effects are selective: small gains appear in meaning and purpose, forgiveness, and sexual satisfaction, with little movement in somatic health, psychological distress, social belonging, or perceived social support. A comparison exposure (+1 hour per week socialising with others) does not reproduce the pattern. We interpret the selective pattern through a prominent cooperative account of religion: gains concentrate in coordination-relevant domains rather than in direct health pathways.