Studies of unconscious plagiarism have reported that people mistakenly include a partner's responses when trying to recall their own (recall-own), but also include their own when recalling a partner's (recall-partner). In a series of simulations we test if a fluency/memory-strength-based explanatory account can accommodate performance in the unconscious plagiarism paradigm. We show that responding on the basis of the typicality of items alone (guessing without memory) cannot account for participants' performance. Adding memory for items generated at study can account for performance in the recall-own but not the recall-partner task, even when we assume that participants in the recall-partner task strategically withhold more fluent items from report. This suggests that judgements of items' memory strength alone cannot explain performance in the unconscious plagiarism paradigm.