Abstract Background The use of molecular techniques to detect malaria parasites has been advocated to improve the accuracy of parasite prevalence estimates, especially in moderate to low endemic settings. Molecular work is time-consuming and costly, thus the effective gains of this technique need to be carefully evaluated. Light microscopy (LM) and rapid diagnostic tests (RDT) are commonly used to detect malaria infection in resource constrained areas, but their limited sensitivity results in underestimation of the proportion of people infected with Plasmodium falciparum . This study aimed to evaluate the extent of missed infections via a community survey in Tanzania, using polymerase chain reaction (PCR) to detect P. falciparum parasites and gametocytes. Methods Three hundred and thirty individuals of all ages from the Kilombero and Ulanga districts (Tanzania) were enrolled in a cross-sectional survey. Finger prick blood samples were collected for parasite detection by RDT, LM and molecular diagnosis using quantitative 18S rRNA PCR and msp2 nPCR. Gametocytes were detected by LM and by amplifying transcripts of the gametocyte-specific marker pfs25 . Results Results from all three diagnostic methods were available for a subset of 226 individuals. Prevalence of P. falciparum was 38% (86/226; 95% CI 31.9–44.4%) by qPCR, 15.9% (36/226; 95% CI 11.1–20.7%) by RDT and 5.8% (13/226; 95% CI 2.69- 8.81%) by LM. qPCR was positive for 72% (26/36) of the RDT-positive samples. Gametocyte prevalence was 10.6% (24/226) by pfs25 -qRT-PCR and 1.2% by LM. Conclusions LM showed the poorest performance, detecting only 15% of P. falciparum parasite carriers identified by PCR. Thus, LM is not a sufficiently accurate technique from which to inform policies and malaria control or elimination efforts. The diagnostic performance of RDT was superior to that of LM. However, it is also insufficient when precise prevalence data are needed for monitoring intervention success or for determining point prevalence rates in countrywide surveillance. Detection of gametocytes by PCR was 10-times more sensitive than by LM. These findings support the need for molecular techniques to accurately estimate the human infectious reservoir and hence the transmission potential in a population.