Ciliates, a eukaryotic clade that is over 1 billion years old, are defined by division of genome function between transcriptionally inactive germline micronuclei and functional somatic macronuclei. To date, most analyses of gene family evolution have been limited to cultivable model lineages (e.g., Tetrahymena, Paramecium, Oxytricha, and Stylonychia). Here, we focus on the uncultivable Karyorelictea and its understudied sister class Heterotrichea, which represent two extremes in genome architecture. Somatic macronuclei within the Karyorelictea are described as nearly diploid, while the Heterotrichea have hyperpolyploid somatic genomes. Previous analyses indicate that genome architecture impacts ciliate gene family evolution as the most diverse and largest gene families are found in lineages with extensively processed somatic genomes (i.e., possessing thousands of gene-sized chromosomes). To further assess ciliate gene family evolution, we analyzed 43 single-cell transcriptomes from 33 ciliate species representing 10 classes. Focusing on conserved eukaryotic genes, we use estimates of transcript diversity as a proxy for the number of paralogs in gene families among four focal clades: Karyorelictea, Heterotrichea, extensive fragmenters (with gene-size somatic chromosomes), and non-extensive fragmenters (with more traditional somatic chromosomes), the latter two within the subphylum Intramacronucleata. Our results show that (i) the Karyorelictea have the lowest average transcript diversity, while Heterotrichea are highest among the four groups; (ii) proteins in Karyorelictea are under the highest functional constraints, and the patterns of selection in ciliates may reflect genome architecture; and (iii) stop codon reassignments vary among members of the Heterotrichea and Spirotrichea but are conserved in other classes.IMPORTANCE To further our understanding of genome evolution in eukaryotes, we assess the relationship between patterns of molecular evolution within gene families and variable genome structures found among ciliates. We combine single-cell transcriptomics with bioinformatic tools, focusing on understudied and uncultivable lineages selected from across the ciliate tree of life. Our analyses show that genome architecture correlates with patterns of protein evolution as lineages with more canonical somatic genomes, such as the class Karyorelictea, have more conserved patterns of molecular evolution compared to other classes. This study showcases the power of single-cell transcriptomics for investigating genome architecture and evolution in uncultivable microbial lineages and provides transcriptomic resources for further research on genome evolution.