Long non-coding RNAs (lncRNAs) represent a diverse class of regulatory loci with roles in development and stress responses throughout all kingdoms of life. LncRNAs, however, remain under-studied in plants compared to animal systems. To address this deficiency, we applied a machine learning prediction tool, Classifying RNA by Ensemble Machine learning Algorithm (CREMA), to analyze RNAseq data from 11 plant species chosen to represent a wide range of evolutionary histories. Transcript sequences of all expressed and/or annotated loci from plants grown in unstressed (control) conditions were assembled and input into CREMA for comparative analyses. On average, 6.4% of the plant transcripts were identified by CREMA as encoding lncRNAs. Gene annotation associated with the transcripts showed that up to 99% of all predicted lncRNAs for Solanum tuberosum and Amborella trichopoda were missing from their reference annotations whereas the reference annotation for the genetic model plant Arabidopsis thaliana contains 96% of all predicted lncRNAs for this species. Thus a reliance on reference annotations for use in lncRNA research in less well-studied plants can be impeded by the near absence of annotations associated with these regulatory transcripts. Moreover, our work using phylogenetic signal analyses suggests that molecular traits of plant lncRNAs display different evolutionary patterns than all other transcripts in plants and have molecular traits that do not follow a classic evolutionary pattern. Specifically, GC content was the only tested trait of lncRNAs with consistently significant and high phylogenetic signal, contrary to high signal in all tested molecular traits for the other transcripts in our tested plant species.