Abstract With the advancement of human stem cell cultures and interest in understanding human embryogenesis, human blastocyst-like structures, or human blastoids, are being developed. Thus far, two different strategies have been taken, generating blastoids from naïve human pluripotent stem cells (hPSC) 1–3 or through reprogramming of adult human somatic cells 4 . All these studies utilize comparative transcriptome analyses to authenticate blastoid cells and identify their counterparts in the human blastocyst 5,6 . However, the validity of comparative transcriptome analysis is critically hinged on including relevant reference data, not only those from targeted cell types but also from potential alternative cell lineages. Thus, we sought to reevaluate the single-cell transcriptome data from the blastoids based on a more comprehensive cellular reference, which includes data from in vitro cultured human 6,7 , non-human primate (NHP, Cynomolgus macaque ) blastocysts 8 , a human stem cell-based post-implantation amniotic sac embryoid (PASE) model 9 , and an in vivo gastrulation-stage human embryo specimen 10 , using four different analysis strategies. Our analyses unequivocally support that blastoids developed by reprogramming adult human somatic cells largely fail to generate cells with a transcriptome profile consistent with the human blastocyst trophectoderm. Instead, cells identified as trophectoderm-like have a transcriptional profile more similar to the amniotic ectoderm in the gastrulating human and NHP embryos. To facilitate cell lineage identifications in human embryo models, we further identified a set of human amniotic ectoderm and trophectoderm markers that could be utilized to distinguish these two lineages. We further built a neural-network based online prediction tool, which accurately discerns the full cellular composition of blastoids.