The coexistence of genetically modified (GM) and non-GM crops is an important economic and political issue in the European Union. We examined the GM content in non-GM maize crops in Spain in 2005. Both the standing crop and the harvest were tested, and the %GM DNA was quantified by real-time polymerase chain reaction. We compared the level of GM as a function of distance from known GM source fields in a 1.2 km2 landscape. The distribution of GM was compared to predictions from previous studies, and good agreement was found. Control and monitoring of adventitious GM presence in non-GM crops can only be achieved by fit-for-purpose sampling and testing schemes. We used a GM dispersal function to simulate non-GM crops in the studied zone and tested the accuracy of five different sampling schemes. Random sampling was found to be the most accurate and least susceptible to bias by GM spatial structure or gradients. Simulations showed that to achieve greater than 95% confidence in a GM labeling decision of a harvest (when treated as a single marketed lot), 34 samples would be needed when the harvest was outside 50% of the GM threshold value. The number of samples required increased rapidly as the harvest approached the GM threshold, implying that accurate labeling when the harvest is within +/-17% of the threshold may not be possible with high confidence.