Aim: A stated advantage of data integration methods to estimate species distributions is flexibility to deal with multiple sources of uncertainty among differing data sets. Two significant sources of uncertainty are high variation in sampling effort across space and observations and reliability associated with data collection protocols or when data was collected. Our goal was to examine how different approaches to address uncertainty affect the predictive performance of integrated models. Location: Pennsylvania, USA Methods: We test this question for models fit using three unique sources of data – which differ in the sampling design used to collect data – to estimate distributions of four bird species. We consider three types of approaches to reduce uncertainty: 1) filtering to reduce which data are included, 2) the form of functions used to account for uncertainty, and 3) how different data sets are integrated into a single estimate. We first focus on accounting for variable effort from community science observations – both related to spatial differences in sampling intensity and the amount of effort given to a single observation record. Results: Sampling effort was best accounted for by conservative filtering of data that included spatial thinning and exclusion of observations that most varied in the effort used to collect the data. Next, we considered how to account for potential false positive detections – due to either misidentification or changes in distributions. We found that treating less reliable data as a covariate, an approach previously suggested for data integration that can greatly speed up model fitting, performed well. Other effective approaches included directly modeling false positive rates and complete exclusion of less reliable data sets. Main Conclusions: We provide insights into best practices in integrated modeling for dealing with uncertainty in integrated models and demonstrate the flexible options available when using integrated models.