We evaluated the use of oral deformities as reliable proxies for determining Batrachochytrium dendrobatidis (Bd) infection in tadpoles of six anuran species of the Atlantic Forest in southeastern Brazil. We examined oral discs of 2156 tadpoles of six species of anurans collected in 2016: Aplastodiscus albosignatus, Boana albopunctata, Boana faber, Scinax hayii, Crossodactylus caramaschii, and Physalaemus cuvieri. Three oral deformities were recognized: lack of keratinization only in upper and/or lower jaw sheaths, lack of keratinization only in upper or lower tooth rows, and both deformities together. A subsample composed of all the individuals possessing oral deformities (N = 195) plus randomly selected individuals without oral deformities (N = 184) were tested for Bd via qPCR. Oral deformities were observed in all six species, but only five were infected with Bd. Since we found that dekeratinization of tooth rows was not associated with the presence of Bd in any of the studied species we used a new proxy (jaw sheaths dekeratinization with or without dekeratinization in tooth rows: JSD-proxy) for Bd detection. Our results showed a nonrandom relationship between Bd infection and JSD-proxy in three species of the family Hylidae. However, the use of JSD-proxy for Bd detection in these species resulted in up to 30.8% false positives and up to 29.3% false negatives. The use of the JSD-proxy in species for which no relationship was found reached 100% of false positives. We conclude that the use of oral dekeratinization as a generalized proxy for Bd detection in tadpoles should not be used as a single diagnosis technique.