Background <p>Thyroid cancer stem cells (CSCs) have been implicated in the recurrence and metastasis of thyroid cancer. This study, therefore, explored the properties of thyroid CSC-like cells, cell–cell interactions, and genes associated with prognosis by analyzing single-cell RNA sequencing (scRNA-seq) data and clinical data.</p> Methods <p>We analyzed two independent scRNA-seq datasets (GSE191288 and GSE250521) using the Seurat R package. Tumor cells were identified based on copy number variations, and stemness^High (CSC-like) and stemness^Low cells were defined using CytoTRACE. Protein–protein interaction (PPI) networks were constructed, after which hub genes were identified using STRING and cytoHubba. Gene ontology analysis was performed using the clusterProfiler R package. CellChat was used for cell–cell interaction analysis. Furthermore, survival analysis was performed using the TCGA database.</p> Results <p>Compared to stemness^Low cells, stemness^High cells showed upregulation of 1,498 (GSE191288) or 1,274 genes (GSE250521), with 109 genes being commonly upregulated in both datasets. Furthermore, hub genes identified from the PPI network constructed from the co-upregulated genes in both two datasets were implicated in angiogenesis. Subsequent cell–cell interaction analysis revealed strong interactions between JAG1 in stemness^High cells and NOTCH1/4 in endothelial cells. Analysis of clinical data showed that thyroid cancer patients with high stemness exhibiting high <i>JAG1</i> expression and that patients with high <i>JAG1</i> expression had significantly poorer prognosis than did those with low <i>JAG1</i> expression.</p> Conclusions <p>This study provides new insights into the gene expression profile of thyroid CSC-like cells and their interactions with cells constituting tumor tissues.</p>