Evidence indicates that following word encoding, false memory is elevated for unstudied words (lures) that share semantic features with studied words. At the same time, the Levels of Processing (LOP) effect suggests that semantic processing in encoding leads to stronger memory. Here, we investigated the interaction of these two effects on false recognition memory across varying levels of study-lure semantic similarity (SS) and LOP at encoding. We first selected lure words; for each, we generated a list of semantically related study words based on a SS measurement from a Natural Language Processing model. Subjects then studied these lists with either a nonsemantic or semantic LOP cover task. Later, subjects made recognition decisions on studied words, unstudied lures, and unstudied new words. Our data revealed a LOP effect; d’ in the semantic LOP condition was significantly higher than d’ in the nonsemantic LOP condition. Further, preliminary results suggest that only in the semantic LOP condition, higher study-lure SS predicted higher false alarm rate for lures. Future analyses will explore these effects with varying levels of study-lure semantic similarity.