Biological conservation of exploited fish species involves characterizing key aspects of their population dynamics using models as tools to estimate their biomass. The Bayesian state-space surplus production model in the open-source stock assessment tool Just Another Bayesian Biomass Assessment (JABBA) was used to assess small yellow croaker stock (Larimichthys polyactis) in the Yellow Sea large marine ecosystem (YSLME). In this study, Catch and the scientific survey catch per unit effort (CPUE) data of the overwintering grounds of small yellow croaker from 1985 to 2020 and three fishery CPUE data series (one original and two reconstructed) from the Chinese Fishery Statistical Yearbook were used to fit JABBA, respectively. The results showed that the trends in biomass obtained from the survey CPUE and from the fishery CPUE contrasted sharply. The independent survey CPUE-based JABBA showed a 54.4 % probability that the current resource status is over exploited (0.65 and 1.06 for B2020/BMSY and F2020/FMSY, respectively), whereas the fishery CPUE-based JABBA showed that the resource is in a healthy or recovering state (0.88–1.32 and 0.40–0.70 for B2020/BMSY and F2020/FMSY, respectively), and the estimates of the resource are overly optimistic. This discrepancy in biomass assessment arises because the fishing effort time series does not adequately reflect the technological advances in fishing vessels and their equipment. Therefore, assessment based on survey CPUE (predicted total allowable catch, TAC = 150,000 tons) is the preferable, more precautionary approach for establishing management reference points and informing management decisions, and estimates generated from the fisheries CPUE-based JABBA model (TAC = 165,000–210,000 tons) should be treated with caution. We found that the two reconstructed fishery CPUEs are more likely to produce model results closer to that of the survey CPUE-based JABBA than the original fishery CPUE. Further review and research on the correction of fishing effort and fishery CPUE in the YSLME is recommended before this data series is used in assessments aimed at biological conservation.