Discourse markers (DMs) are linguistic elements that play a vital role in facilitating coherence and structuring spoken interaction, particularly in spontaneous communication where speakers must organize their ideas in real time. This study examines their types and functions in athlete-hosted podcast conversations. Using a qualitative descriptive design with document analysis, transcribed podcast data were analyzed to identify discourse marker types and explore their functions within spontaneous speech. A total of 2,013 discourse markers were identified. Six types emerged, each serving distinct pragmatic functions. Discourse connectives (39.90%) such as and, but, also, and then maintained coherence by linking ideas. Markers of information and participation (40.58%) like you know, I think, I mean, yeah, like, and actually expressed stance, clarified utterances, and engaged listeners. Information management markers (1.69%) such as okay and well indicated topic shifts or cognitive processing. Cause-result markers (8.44%) such as because and so explained reasoning and outcomes. Response markers (5.81%) including yeah and woah conveyed agreement or emotional reaction. Temporal/modal adverbs (3.57%) such as now, tomorrow, and sometimes marked time, while probably and maybe expressed uncertainty. This analysis confirms that each type of discourse marker contributes uniquely to achieving communicative goals in spontaneous speech. These findings demonstrate that athlete-hosted podcasts provide authentic and pragmatically rich spoken data, making them valuable for discourse analysis and language teaching. Future researchers are recommended to examine discourse marker usage in different context such as interview, tv shows, and debate.