The gold standard method in the diagnosis of SARS-CoV-2 infection is the detection of viral RNA in the nasopharyngeal sample by RT-PCR. Recently, saliva samples have been suggested as an alternative sample. In the present study, we aimed to compare RT-PCR results in nasopharyngeal, oro-nasopharyngeal and saliva samples of COVID-19 patients. 98 of 200 patients were positive in RT-PCR analysis performed before the hospitalization. On day 0, at least one sample was positive in 67 % of 98 patients. The positivity rate was 83 % for both oronasopharyngeal and nasopharyngeal samples, while it was 63 % for saliva samples (p < 0.001). On day 5, RTPCR was performed in 59 patients, 34 % had at least one positive result. The positivity rate was 55 % for both saliva and nasopharyngeal samples, while it was 60 % for oro-nasopharyngeal samples. Our study shows that the sampling saliva does not increase the sensitivity of RT-PCR tests at the early stages of infection. However, on the 5th day, viral RNA detection rates in saliva were similar to nasopharyngeal and oro-nasopharyngeal samples. In conclusion, we suggest that, in patients receiving treatment, RT-PCR in saliva, in addition to the standard samples, is important to determine the isolation period and control transmission.