Deliberative mini-publics are currently being employed throughout the world, and it has been well documented that women are less likely to exercise their voice in deliberation. Yet the advent of digital platforms that host discussions online could mitigate this gendergap, and make it easier to test theories about the design of deliberative institutions. Increasing descriptive representation — the number of women in a discussion group — could potentially improve women’s frequency of participation and quality of experience; however, we know less about how these dynamics play out in deliberative settings that take place entirely online. Here, we leverage unique data from the 2023 Meta Community Forum on AI Chatbots, which brought together 1,541 people in a Deliberative Polling exercise that took place on a digital platform. Thanks to random assignment, the gendercomposition of each discussion room varies. We exploit the group composition to estimate how the number of female members in each group affects participation (as measured by speaking frequency and activity during the session), as well as self-reported measures of efficacy and satisfaction with deliberative processes. We find some descriptive evidence of a gender gap, but little to no effects of room composition on participation or evaluation; this has surprising implications for the design of digital deliberation.