Streptococcus pneumoniae causes invasive diseases of significant public health concern, such as meningitis. The culture of cerebrospinal fluid (CSF) samples, the standard technique for meningitis diagnoses, is not always positive. Consequently, meaningful information about the etiological agent is lost, which can compromise effective epidemiological surveillance and the improvement of immunization policies. This study aims to standardize a method to genotype pneumococcus in the CSF samples which could mitigate the absence of isolated strains, and also evaluate the prediction of this assay. We applied eight multiplex PCR (mPCR) assays to CSF samples paired with the Quellung reaction applied to the isolated strains. We also compared different master mix kits in the mPCR. Moreover, a retrospective study was conducted with CSF samples considered pneumococcus positive due to the presence of the lytA gene. Results showed that genotyping by the mPCR correlated 100% with the Quellung reaction, and genotyping was dependent on the master mix applied. In the retrospective study (2014–2020), 73.4% were successfully genotyped. The analyses of the receiver operating characteristic curve showed that the cycle threshold (Ct value) around 30 for the lytA gene had a 75% positive chance of successful genotyping, whereas with a Ct value > 35, the chance was 12.5%. Finally, we observed that genotype 19A was prevalent in the period (12%), information unknown until now due to the lack of isolated strains. Therefore, the mPCR of CSF samples can efficiently predict S. pneumoniae serotypes, especially in the absence of isolated strains, which can be a great tool for pneumococcal serotype surveillance.