<p>This quantitative study examines the impact and predictive power of AI-assisted spoken English practice and affective factors (self-confidence, risk-taking, L2 speaking anxiety, grit, L2 speaking enjoyment) on L2 willingness to communicate (WTC), after controlling for demographic characteristics (gender, age, major). The study involved 354 Chinese undergraduate students who learned English as a foreign language. Results from hierarchical regression analysis showed that affective factors collectively explained 83.9% of the variance in L2 WTC, with speaking enjoyment emerging as the strongest predictor, followed by self-confidence and grit; risk-taking and speaking anxiety were non-significant. After controlling for demographics, AI usage had a significant but weak positive impact on L2 WTC (explaining 6.5% of variance), and its effect remained significant but diminished when affective factors were included. Path analysis confirmed a mediating model where self-confidence and grit influence L2 WTC through risk-taking (weakly) and speaking enjoyment (strongly). Multi-group comparisons indicated that this model was consistent across AI users and non-users. These findings establish the primacy of positive affective factors as the primary drivers of L2 WTC and reveal that AI-mediated practice, while secondary, enhances WTC without restructuring its core affective pathways. Consequently, this study calls for pedagogies that centrally foster these affective factors, strategically leveraging AI as a complementary scaffold.</p>