International plant breeding efforts aim to help address global challenges such as food insecurity, climate change and micronutrient deficiencies, but how much do we know about where new varieties are being used at scale in farmers’ fields? We use farmer-reported survey data and DNA fingerprinting data from samples of plant tissue that were collected from the same farmers’ fields, from 16 empirical studies across nine crops and ten countries to test commonly used methods used to generate adoption estimates. Taking the DNA fingerprinting results as the benchmark, we show that farmers both misclassify (improved vs local varieties) and misidentify (by name) their varieties at substantial rates. These measurement errors are non-classical and thus cloud understanding of the true nature of adoption.