Research over the past 20 years has suggested that our intuitive sense of number--the Approximate Number System (ANS)--is associated with individual differences in symbolic math performance. The mechanism supporting this relationship, however, remains unknown. Here, we test whether the ANS contributes to how well adult observers judge the "direction" and "magnitude" of symbolic math equation errors. We developed a novel task in which participants view symbolic equations with incorrect answers (e.g., 47 + 21 = 102), and indicate whether the provided answer was too high or too low. By varying the ratio between the correct and the provided answers, we measured individual differences in how well participants detect the "magnitude" and "direction" of symbolic equation errors. We find that individual differences in equation error detection were uniquely predicted by ANS acuity--that is, the precision of each participant's intuitive number representations--even when controlling for differences in surface area perception, working memory span, and operational span. This suggests that the ANS can act as a unique source of error detection variability for formal mathematics, providing a plausible mechanism for how our universally shared number sense might link with human-specific symbolic math abilities.