Programmable thermostats are generally sold as energy-saving devices controlling heating and cooling systems, but can lead to energy waste when not operated as designed by the manufacturers. We utilized Amazon Mechanical Turk, an online crowdsourcing service, to investigate thermostat settings and behavior in households. We posted a survey and paid respondents to upload pictures of their thermostats to verify self-reported data. About 40% of programmable thermostat owners did not use programming features and 33% had programming features overridden. Respondents demonstrated numerous misconceptions about how thermostats control home energy use. Moreover, we found that 57% of households were occupied nearly all the time, limiting the potential energy savings. The study revealed flaws in selfreported data, when collected solely from traditional surveys, which raises concerns about the validity of current thermostat-related research using such data. “Ground truth” temperature data could now be available in homes with Internet-connected thermostats. Online crowdsourcing platforms emerge as valuable tools for collecting information that would be difficult or expensive to obtain through other means. Advantages over traditional surveys include low-cost, rapid design–implementation–result cycle, access to diverse population, use of multimedia. Crowdsourcing is more effective than alternative online tools due to easier recruitment process and respondents’ reputation system.