What AI cannot sense: a situated reading of olive oil-based cosmetic care

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This article develops a critical reflection on the growing integration of artificial intelligence (AI) in the cosmetics industry, drawing on the philosophy of technology, the sociology of situated knowledge, and an analysis of the heritage of body care practices. Using the emblematic example of the traditional use of olive oil in skin care practices in Morocco, it questions how AI systems, under the guise of personalization, reproduce standardized aesthetic norms and contribute to the erasure of perceptual, affective, and vernacular knowledge, while revealing algorithmic biases. Beauty is understood as a form of knowledge rooted in cultures, carried by knowledge transmitted through attention to the body, narratives, and sociocultural contexts. This perspective is part of an epistemology of the sensory and the relational, which highlights the roots of these practices in local forms of validation and regulation, far removed from the logic of abstraction inherent in contemporary technologies. Whereas AI operates by extracting pixels, textures, or hues, traditional cosmetics mobilize tactile, memory-based, and relational intelligence. To put these claims to empirical test, a prompting experiment was conducted with four large language models, asking them to advise on Moroccan olive oil-based care practices with progressively richer contextual information. The results reveal that even when models produce culturally informed responses, they remain structurally incapable of accessing the embodied, relational, and sensory dimensions of these practices, confirming that the gap between algorithmic knowing and situated knowledge is ontological rather than technical. We advocate for a more equitable, localized AI in cosmetics that is attentive to local contexts, co-constructed with communities, respectful of ecologies, and open to invisible voices. Rather than attempting to translate vernacular knowledge into digital protocols, we propose the concept of fragile bridges: AI systems designed to document the conditions under which situated practices emerge, to integrate contestability mechanisms, and to remain transparent about what algorithmic formalization cannot carry. We propose three conceptual shifts, namely reading cosmetics as embodied intelligence; recognizing locally rooted feminine knowledge as a form of everyday data science; and conducting a systemic critique of the AI washing logic that structures beauty tech.