Author networks play a key role in doing science. Developing networks is critical for career advancement in a wide variety of ways, and differences in networks may be a core reason for persistence of implicit gender bias. Combining the AGU Fall Meeting abstracts from 2014-2018 with self-identified AGU member data on birth year and gender provides a large database of more than 400,000 unique co-author interactions that we use to examine author networks by age, gender, and country. Age data are necessary to disambiguate the effect that a historic lack of women in the Earth and space science. The data show that women’s networks are closer to those expected from the age-gender distribution of the overall membership; whereas networks of men include more men than expected, although women are also interacting with men of similar age more than expected from the membership. Women’s networks are also less international than their male colleagues in most age cohorts. These differences start in the youngest age cohort. These data indicate that addressing implicit bias requires efforts at encouraging and developing more balanced author networks, particularly in early-career scientists. Recent work suggest that this will also improve science outputs.