Understanding the prevalence of clinically relevant pharmacogenetic variants using large unselected populations is critical for gauging the potential clinical impact of widespread preemptive pharmacogenetic testing. To this end, we assessed the frequencies and ethnic distribution of the three most common CYP2C19 alleles (*2, *3, and *17) in 2.29 million direct-to-consumer genetics research participants (23andMe, Inc. Sunnyvale, CA). The overall frequencies of *2, *3, and *17 were 15.2%, 0.3%, and 20.4%, respectively, but varied by ethnicity. The most common variant diplotypes were *1/*17 at 26% and *1/*2 at 19.4%. The less common *2/*17, *17/*17 and *2/*2 genotypes occurred at 6.0%, 4.4%, and 2.5%, respectively. Overall, 58.3% of participants had at least one increased-function or no-function CYP2C19 allele. To better understand how this high frequency might impact a real patient population, we examined the prescription rates (Rx) of high-pharmacogenetic-risk medications metabolized by CYP2C19 using the University of California at San Francisco (UCSF) health system's anonymized database of over 1.25 million patients. Between 2012 and 2019, a total of 151,068 UCSF patients (15.8%) representing five self-reported ethnicities were prescribed one or more high-pharmacogenetic-risk CYP2C19 medications: proton pump inhibitors (145,243 Rx), three SSRI antidepressants (54,463 Rx), clopidogrel (14,376 Rx), and voriconazole (2,303 Rx).