Large language models (LLMs) are transforming student learning, mirroring how search engines previously disrupted traditional information sources. This study compares these technologies through a within-subjects study where participants learned topics using both Google and ChatGPT, followed by interviews exploring their experiences. Our analysis reveals nuanced insights into students’ strategic tool selection processes, showing how contextual factors influence when and why students prefer LLMs over search engines, with particular attention to emergent trust heuristics, epistemic challenges, and evolving notions of authorship and verification–with implications for education stakeholders navigating this technological shift.