Speaking anxiety and fear of poor grades make it difficult for EFL students to speak well, leading them to seek alternative learning methods that offer safe practice or real-life communication. This qualitative case study examines how 10 second-semester English language students at Semarang State University felt about practicing their English using LearnAI (AI-based interaction) and HelloTalk (human-based interaction). Braun and Clarke's thematic analysis was used to examine data derived from semi-structured interviews and descriptive questionnaires. The results indicate that students carefully planned the sequence of their applications based on their emotional readiness: As a place to practice, LearnAI helped students feel more confident in their technical skills and reduced anxiety about public speaking by providing a judgment-free space. On the other hand, HelloTalk was a place for real-life social interaction that helped students improve their pragmatic competence. Three distinct types of anxiety were identified: language-based anxiety, speech-based anxiety, and safety-based anxiety. This research supports a model that combines Krashen's (1982) Affective Filter Hypothesis with Vygotsky's (1978) Sociocultural Theory. This model posits that emotional safety precedes actual social engagement. As a teaching method, teachers should use a step-by-step approach, starting with organized, low-stress practice with AI-based tools, then moving on to real conversations with native speakers. This way, students can improve their technical proficiency and comfort in communicating.
Copyrights © 2026