Amidst social distancing, quarantines, and everyday disruptions caused by the COVID-19 pandemic, users’ heightened activity on online social media has provided enhanced opportunities for self-disclosure. We study the incidence and the evolution of self-disclosure temporally as important events unfold throughout the pandemic’s timeline. Using a BERT-based supervised learning approach, we label a dataset of over 31 million COVID-19 related tweets for self-disclosure. We map users’ self-disclosure patterns, characterize personal revelations, and examine users’ disclosures within evolving reply networks. We employ natural language processing models and social network analyses to investigate self-disclosure patterns in users’ interaction networks as they seek social connectedness and focused conversations during COVID-19 pandemic. Our analyses show heightened self-disclosure levels in tweets following the World Health Organization’s declaration of pandemic worldwide on March 11, 2020. We disentangle network-level patterns of self-disclosure and show how self-disclosure characterizes temporally persistent social connections. We argue that in pursuit of social rewards users intentionally self-disclose and associate with similarly disclosing users. Finally, our work illustrates that in this pursuit users may disclose intimate personal health information such as personal ailments and underlying conditions which pose privacy risks.