Occupations are defined as categories of work containing similar jobs and tasks. But how similar are these jobs and tasks? To what extent do people in the same occupation describe their work in cohesive ways on household surveys? A limited body of prior research has found that occupational categories are highly internally varied and becoming more so over time, with implications for how we understand the meaning of occupations and how occupations relate to inequality. However, these findings are based on relatively small samples, with limited coverage of detailed occupational categories. To address this limitation, I construct an index of occupational similarity based on write-in textual data on job names and tasks from the 2011 to 2021 American Community Survey, which has over a million observations per year. Findings indicate differential levels of occupational similarity cross-sectionally, but little change in similarity over the period studied. Similarity is highly sensitive to changes in the survey questionnaire, reinforcing the importance of data collection methods for occupation. Replicating prior research, this paper finds that occupational similarity is highly correlated with income, education, and demographic composition. It also finds that similarity predicts within-occupation income inequality. These results show that the similarity index is a useful tool for measurement and interpretation of occupations.