Primary Progressive Aphasia (PPA) is a neurocognitive disorder ascribed to cortical atrophy impacting language abilities. It is widely classified into three main variants, logopenic PPA (lvPPA), the semantic variant of PPA (svPPA), and the non-fluent PPA (nfvPPA), showing different impairment patterns across variants. However, in the early phases of PPA, it is not always easy to dissociate different PPA variants and distinguish PPA from other neurodegenerative disorders. One characteristic language symptom that seems to be a distinguishing factor of PPA, especially the logopenic variant, is impaired sentence repetition. Nonetheless, studies examining sentence repetition in PPA, and Alzheimer's disease (AD) more broadly, have resulted in mixed findings. To better understand the working memory-intensive nature of sentence repetition deficits, we designed a sentence repetition span task. We seek to understand (i) whether three diagnostic groups (lvPPA, svPPA, and AD) encounter greater sentence repetition difficulties than the controls, and (ii) whether using a span task design, in which the number of content words increases as the span length increases, would help dissociate PPA variants from AD type dementia. In this study, we administered a sentence repetition span task to four groups of French-speaking individuals with lvPPA (n = 14), svPPA (n = 5), and with AD (n = 13), and their age-matched healthy controls (n = 61). The results showed that all three diagnostic groups (lvPPA, svPPA, and AD) performed equally poorly compared to the controls on the repetition span task virtually in all measures (i.e., sentence span, the number of content words, and the number of omission and substitution errors). One intriguing finding was that the lvPPA group produced an exalted number of phonological errors during repeating sentences, while this type of error was somewhat moderate in the svPPA group and only minimal in the AD group. We conclude that the sentence repetition difficulty in PPA and AD should be modulated by working memory capacity, as our participants undoubtedly demonstrated greater difficulty as the span length increased. However, we note that working memory-intensive sentence repetition deficits based on the number of content words might not reveal critical diagnostic differences between the neurodegenerative groups.