Priority users (e.g., verified profiles on Twitter) are social media users whose content is promoted by recommendation algorithms. However, the impact of this heterogeneous user influence on opinion dynamics, such as polarization phenomena, is unknown. We conduct a computational mechanistic investigation of such consequences in a stylized setting. First, we allow priority users, whose content has greater reach (similar to algorithmic boosting), into an opinion model on adaptive networks. Then, to exploit this gain in influence, we incorporate stubborn user behavior, i.e., zealot users who remain committed to opinions throughout the dynamics. Using a novel measure of echo chamber formation, we find that prioritizing users can inadvertently reduce polarization if they post according to the same rule but sharpen echo chamber formation if they behave heterogeneously. Moreover, we show that a minority of extremist ideologues (i.e., users who are both stubborn and priority) can push the system into a transition from consensus to polarization with echo chambers. Our findings imply that the implementation of the platform’s prioritization policy should be carefully monitored in order to ensure there is no abuse of users with extra influence.