Policymakers often promote the importance of STEM jobs but are concerned about the underrepresentation of women and minorities in these jobs. However, there is no agreed-upon definition of STEM jobs. I use occupation task data from O*Net to analyze the STEM task content of occupations, drawing several conclusions. First, there is no clear, robust definition of STEM occupations, even when using task data. The occupations included are highly sensitive to the cut-offs and methods used. Second, there are a number of occupations that should clearly be considered STEM by task content but are typically not, including nurse practitioners, pharmacists, and economists. Third, the gender gap in STEM jobs depends heavily on how one defines STEM. One traditional definition shows that STEM jobs are 76% male, but most task-based definitions show gender gaps only half as large (62-65% male). Racial gaps in STEM and the earnings premium for STEM occupations (35-43%) are fairly stable across definitions. The results imply that policies promoting traditionally-defined STEM jobs can unnecessarily exclude women and draw workers away from other important occupations.