Objective: Web-based relationship interventions have been shown to improve relationship functioning for low-income couples, with gains maintained up through 12 months (Doss et al., 2020; Roddy et al., 2020). However, little is known about mechanisms of improvements and maintenance for these programs. Method: A sample of 742 low-income couples (M-age = 33; 55% non-Hispanic White; 52% married; Mdn annual household income = $27,000) were randomly assigned to the OurRelationship (OR) program, the ePREP program, or a waitlist control group. Couples were repeatedly assessed over 6 months on two relationship outcomes (satisfaction and breakup potential) and five relationship mechanisms (positive and negative communication, relationship problem intensity and confidence, and emotional support) with self-report measures. Results: Intervention couples, compared with the control couples, experienced significantly greater improvements in all mechanisms by the end of the program (Mean vertical bar d vertical bar = 0.58). Those with greater improvements in mechanisms also reported greater improvements in both relationship outcomes. From post- to 4-months follow-up, only prepost improvements in negative communication significantly predicted maintenance of gains in satisfaction. In contrast, prepost improvements in all mechanisms significantly predicted maintenance of reductions in breakup potential, with the largest effect via improvements in emotional support. Conclusions: These results indicate that both programs improve targeted mechanisms during the intervention period and support the role of these constructs as mechanisms of improvements in broader relationship functioning. Moreover, findings on mechanisms of maintenance highlighted the importance of including an emphasis on communication and emotional support to better maintain program gains.