Autism is a neurodevelopmental condition with early onset, usually entailing language differences compared to neurotypical peers. Females are four times less likely than males to be diagnosed with autism, and the language features associated with this condition are less frequent in females than in males. However, the developmental mechanisms underlying these sex differences remain unclear. In neurotypical populations, sex differences in language development are also observable from early on, with females outperforming males. One mechanism underlying these sex differences may be early differences in selective attention to talking faces. During the first year, more mouth-looking generally predicts better language development, but sex differences exist. Female infants look at the mouth of a talking face more than males without penalizing looking to the eyes, and reduced mouth-looking in early infancy relates to better vocabulary in toddlerhood only in females. In this hypothesis and theory article, we propose that unique female gaze patterns to the mouth may constitute an early female-specific candidate marker that acts as a protective marker for language development also in autism. Since autism is highly heritable, investigating infants at elevated likelihood for autism offers the opportunity to search for sex-specific markers operating early in life before autistic features and language differences emerge. We argue that, as in neurotypical female infants, mouth-looking may also protect female infants-at-elevated-likelihood-for-autism population from potential later differences in language skills. If so, then sex-specific early behavioral markers, potentially acting as protective markers of language, may compensate for some genetic risk markers affecting this population. Here we gather evidence from neurotypical infants and those with elevated likelihood of autism to uncover why biological sex, the development of selective attention to the mouth, and language acquisition could be intimately related in both populations. We also propose hypotheses regarding potential sex-differentiated neurodevelopmental pathways. We end discussing future research challenges: how generalizable mouth-looking could be as a potential female-specific early language marker across contexts (experimental vs. real life), countries, and developmental time. Ultimately, we aim to target a novel protective candidate of language acquisition, informing tailored interventions that consider sex as an important source of individual variability.