<p>Science is often portrayed as a producer of reliable knowledge, yet scientists routinely express caution, acknowledging doubt and complexity. This is true especially in emerging fields marked by deep uncertainty such as astrobiology, which explores difficult questions about the origin, evolution, and distribution of life on Earth and beyond. Mapping the expression of uncertainty in publications helps refine our understanding of how scientific knowledge is constructed and highlights domains at the frontiers of the unknown. This study investigates the presence, distribution, and context of uncertainty in a corpus comprising the complete set of research articles (<i>N</i> = 3698) published in the three leading astrobiology journals, from their inception through 2020. Using a linguistically grounded method to detect uncertainty markers, combined with Latent Dirichlet Allocation (LDA) topic modeling, we analyze how uncertainty varies across subfields and over time. Our findings reveal a strong link between uncertainty and the rhetorical organization of scientific writing, with uncertainty clustering in introductions, discussions, and conclusions—sections that function as rhetorical spaces for doubt and reflection. The expression of uncertainty also significantly varies across research topics, being highest in speculative areas such as the nature of life or extraterrestrial intelligence, and lowest in experimentation driven fields. Finally, an analysis of epistemic markers shows that highly uncertain topics tend to use fewer knowledge-related terms, whereas empirically grounded domains display richer epistemic vocabularies indicating that uncertainty reflects not only research content but also differences in scientific practice and epistemic framings.</p>