This study investigates the use of lexical bundles in the listening section of high school English assessments. To better understand authentic language use, facilitate learners’ vocabulary development, and inform effective pedagogical strategies and test design, this study analyzed frequency, structural, and functional differences among lexical bundles in English listening sections of the CSAT, Academic Achievement Tests, Mock Exams, and a native English reference corpus (i.e., the American National Corpus). Results indicated that listening assessments contained more lexical bundles than the ANC. Structural analysis showed that phrasal bundles were the most frequently used in both corpora, with verb phrases being the second most common in listening assessments, whereas noun phrases were the second most common in the ANC. Functional analysis revealed that stance bundles were most frequently employed in both corpora. Special conversational functions were the second most frequent in listening assessments, while they were the least frequent in the ANC. Additionally, referential expressions and discourse organizers occurred less frequently than other functions. Both corpora thus exhibited characteristics of a spoken register. These findings underscore the importance of context-based learning that focuses on multi-word expressions and lexical bundles to enhance learners’ vocabulary skills.