Background Adults are being vaccinated against SARS-CoV-2 worldwide, but the longitudinal protection of these vaccines is uncertain, given the ongoing appearance of SARS-CoV-2 variants. Children remain largely unvaccinated and are susceptible to infection, with studies reporting that they actively transmit the virus even when asymptomatic, thus affecting the community. Methods We investigated if saliva is an effective sample for detecting SARS-CoV-2 RNA and antibodies in children, and associated viral RNA levels to infectivity. For that, we used a saliva-based SARS-CoV-2 RT-qPCR test, preceded or not by RNA extraction, in 85 children aged 10 years and under, admitted to the hospital regardless of COVID-19 symptomatology. Amongst these, 29 (63.0%) presented at least one COVID-19 symptom, 46 (54.1%) were positive for SARS-CoV-2 infection, 28 (32.9%) were under the age of 1, and the mean (SD) age was 3.8 (3.4) years. Saliva samples were collected up to 48 h after a nasopharyngeal swab-RT-qPCR test. Results In children aged 10 years and under, the sensitivity, specificity, and accuracy of saliva-RT-qPCR tests compared to NP swab-RT-qPCR were, respectively, 84.8% (71.8%–92.4%), 100% (91.0%–100%), and 91.8% (84.0%–96.6%) with RNA extraction, and 81.8% (68.0%–90.5%), 100% (91.0%–100%), and 90.4% (82.1%–95.0%) without RNA extraction. Rescue of infectious particles from saliva was limited to CT values below 26. In addition, we found significant IgM positive responses to SARS-CoV-2 in children positive for SARS-CoV-2 by NP swab and negative by saliva compared to other groups, indicating late infection onset (>7–10 days). Conclusions Saliva is a suitable sample type for diagnosing children aged 10 years and under, including infants aged <1 year, even bypassing RNA extraction methods. Importantly, the detected viral RNA levels were significantly above the infectivity threshold in several samples. Further investigation is required to correlate SARS-CoV-2 RNA levels to viral transmission.