Voice assistants like Amazon's Alexa, Google's Assistant, or Apple's Siri,\nhave become the primary (voice) interface in smart speakers that can be found\nin millions of households. For privacy reasons, these speakers analyze every\nsound in their environment for their respective wake word like ''Alexa'' or\n''Hey Siri,'' before uploading the audio stream to the cloud for further\nprocessing. Previous work reported on the inaccurate wake word detection, which\ncan be tricked using similar words or sounds like ''cocaine noodles'' instead\nof ''OK Google.''\n In this paper, we perform a comprehensive analysis of such accidental\ntriggers, i.,e., sounds that should not have triggered the voice assistant, but\ndid. More specifically, we automate the process of finding accidental triggers\nand measure their prevalence across 11 smart speakers from 8 different\nmanufacturers using everyday media such as TV shows, news, and other kinds of\naudio datasets. To systematically detect accidental triggers, we describe a\nmethod to artificially craft such triggers using a pronouncing dictionary and a\nweighted, phone-based Levenshtein distance. In total, we have found hundreds of\naccidental triggers. Moreover, we explore potential gender and language biases\nand analyze the reproducibility. Finally, we discuss the resulting privacy\nimplications of accidental triggers and explore countermeasures to reduce and\nlimit their impact on users' privacy. To foster additional research on these\nsounds that mislead machine learning models, we publish a dataset of more than\n1000 verified triggers as a research artifact.\n