Structured RNAs have emerged as a major component of cellular regulatory systems, but their mechanism of action is often poorly understood. Riboswitches are structured RNAs that allosterically regulate gene expression through any of several different mechanisms. In vitro approaches to characterizing mechanism are costly, low-throughput, and must be repeated for each individual riboswitch locus of interest. Bioinformatic methods promise higher throughput; despite robust computational identification of riboswitches, however, computational classification of riboswitch mechanism has so far been both model-bound, relying on identification of sequence motifs known to be required for specific models of riboswitch activity, and empirically untested, with predictions far outpacing biological validation. Here, we introduce TaRTLEt (Transcriptionally-active Riboswitch Tracer Leveraging Edge deTection), a new high-throughput tool that recovers in vivo patterns of riboswitch-mediated transcription termination from paired-end RNA-seq data using edge detection methods. TaRTLEt successfully extracts transcription termination signals despite numerous sources of biological and technical noise. We tested the effectiveness of TaRTLEt on riboswitches identified from a wide range of sequenced bacterial taxa by utilizing publicly available paired-end RNA-seq readsets, finding broad agreement with previously published in vitro characterization results. In addition, we use TaRTLEt to infer the in vivo regulatory mechanism of uncharacterized riboswitch loci from existing public data. TaRTLEt is available on GitHub and can be applied to paired-end RNA-seq datasets from isolates or complex communities.