Not sure where we all seemed to hear that one (I'm guessing some movie made it a big plot point), but there's at least one actual study that disagrees:
>In short, all three studies provided no evidence to support the notion that the patterns of eye-movements promoted by many NLP practitioners aid lie detection. This is in line with findings from a considerable amount of previous work showing that facial clues (including eye movements) are poor indicators of deception [1]
https://journals.plos.org/plosone/article?id=10.1371/journal...