In the run up to the 2016 U.S. presidential election, two scandals were front and center in the American news, social media, and across the internet. “Bridgegate” was based on verifiable events related to the closure of several lanes leading to the George Washington bridge in Ft. Lee, New Jersey. “Pizzagate” was based entirely on an ideologically-driven fiction that was presented through a series of stories told as true, in which high ranking members of the Democratic establishment were alleged to be involved in a child-traficking ring operating out of the basement of a Washington DC pizza parlor. This type of informal, yet ideological, storytelling constitutes a large part of everyday interaction, both in the online and offline worlds. In this work, we propose a three-level generative model of everyday storytelling. This model makes use of the “deep structure” models of earlier studies, yet accommodates the incomplete and noisy storytelling that characterizes most online and face-to-face interactions, by inserting an intermediary meso-level. This intermediary level allows us to uncover the emergence of stable narrative frameworks that reveal the dynamic relationships between actants, and trace the shifts in that framework caused by the observable aspects of storytelling. Using relatively straightforward computational methods, we are able to derive for any domain the narrative framework and match observed stories to that framework. We base this work on three main case studies: legends of witchcraft from 19th century Denmark, stories of vaccine hesitancy among American parents over the past decade, and the Pizzagate conspiracy.