Long non-coding RNAs (lncRNAs) are a diverse class of transcribed RNA molecules that are, by definition, at least 200 nucleotides in length and do not encode proteins. While tens of thousands of lncRNAs have been annotated in mammalian genomes, only a small fraction of these have been functionally characterized to date. At the bulk tissue level, lncRNAs are relatively lowly expressed compared to protein-coding genes, often to the point that they fall below the limit of detection of RNA sequencing (RNA-seq) assays. This has led some observers to conclude that lncRNAs largely represent transcriptional ‘noise’ that results from low frequency or stochastic transcription/splicing events. However, such bulk tissue studies measure the average expression of a given gene across many thousands (or millions) of individual cells; thus, it remains an open possibility that lncRNAs are relatively highly expressed in a small fraction of cells (or restricted a particular cell type). Here, we combine targeted RNA capture, single-cell RNA-seq and long-read isoform sequencing to profile lncRNA expression in mouse cortex at an unprecedented level of transcriptomic resolution. We find that, while ~85% of all targeted lncRNAs are expressed in mouse cortex, only a small fraction of these (~1%) are detected in any given individual cell. We also show that certain cell types express specific lncRNA genes or isoforms exclusively, providing new evidence for their functional importance. Furthermore, we show that while lncRNA exons are universally alternatively spliced at the bulk tissue level, their inclusion rates approach that of protein-coding exons at the single-cell level. This study provides the first comprehensive single-cell, isoform-level map of lncRNA expression in the mouse brain, and suggests that lncRNA expression is much more precise and cell-specific than previously assumed.