OBJECTIVE: The objective of this study was to investigate the psychometric properties of the Oral Health Literacy Assessment in Spanish (OHLA-S) for the Brazilian-Portuguese language using robust analysis and with the results disclose possibilities to develop a shorter and more valid instrument. METHODS: OHLA-S is an oral health literacy instrument comprising a word recognition section and a comprehension section. It consists of 24 dental words. It was translated into the Brazilian-Portuguese language (OHLA-B) and its psychometric properties were evaluated in a random sample of 250 adults aged 20-59 years. To assess the dimensionality and factor structure were tested by means of Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA). Reliability was assessed using two indicators: Cronbach's alpha and McDonald's Omega. RESULTS: EFA and CFA demonstrated that the OHLA-B with 24 items did not present an adequate adjustment of the model, compromising its validity. In addition, reliability values at 0.50 for Cronbach's alpha and 0.67 for McDonald's omega were below the minimum acceptable rate of 0.70. As no support was found for the original structure, we decided to proceed with the withdrawal of individual items and successive reanalysis of the model until the indicators were adjusted in a shorter instrument. A new structure with 15 items produced an instrument with two dimensions and a better goodness of fit than the original instrument. The Alpha and Omega reliability index values increased to 0.83 and 0.80, respectively, and all scores were better in the OLHA-B with 15 items than in the instrument with 24 items. CONCLUSION: OLHA-B with the original structure composed by 24 items did not show acceptable construct validity. The shorter version with 15 items showed more promising results for assessing oral health literacy levels in the Brazilian population.