Studies of improved seed adoption in developing countries almost always draw from household surveys and are premised on the assumption that farmers are able to self-report their use of improved seed varieties. However, recent studies suggest that farmers’ reports of the seed varieties planted, or even whether seed is local or improved, are sometimes inconsistent with the results of DNA fingerprinting of farmers' crops. We use household survey data from Tanzania to test the alignment between farmer-reported and DNA-identified maize seed types planted in fields. In the sample, 70% of maize seed observations are correctly reported as local or improved, while 16% are type I errors (falsely reported as improved) and 14% are type II errors (falsely reported as local). Type I errors are more likely to have been sourced from other farmers, rather than formal channels. An analysis of input use, including seed, fertilizer, and labor allocations, reveals that farmers tend to treat improved maize differently, depending on whether they correctly perceive it as improved. This suggests that errors in farmers' seed type awareness may translate into suboptimal management practices. In econometric analysis, the measured yield benefit of improved seed use is smaller in magnitude with a DNA-derived categorization, as compared with farmer reports. The greatest yield benefit is with correctly identified improved seed. This indicates that investments in farmers' access to information, seed labeling, and seed system oversight are needed to complement investments in seed variety development.