Generative AI Engagement and Perceived Nutrition-Oriented Feeding Practices Among Urban Indonesian Mothers: Mixed Methods Study.

Abstract

Background: The Asia-Pacific region faces an escalating double burden of malnutrition, with childhood obesity now affecting more than 113 million children aged 0-19 years. A demographic of highly educated, digitally literate urban mothers-referred to in this study using the descriptive label "Alpha Mother"-increasingly draws on generative AI as a source of toddler nutrition information. Yet, whether this shift empowers or undermines quality feeding practices remains empirically untested.

Objective: This study examined how generative artificial intelligence (AI) engagement is associated with perceived nutrition-oriented feeding practices among urban Alpha Mothers in Indonesia, with cognitive empowerment and parental self-compassion as intervening variables, and eHealth self-efficacy as a moderator. The study contextually refined 5 AI engagement dimensions, organized through an AI parenting nutrition alignment approach.

Methods: An exploratory sequential mixed methods design was used. Phase 1 comprised semistructured interviews with 10 Alpha Mothers and used hybrid deductive-inductive thematic analysis to contextually refine 5 literature-informed AI engagement constructs: AI algorithmic trust, AI health information quality, AI personalization fit perception, AI information-seeking intensity, and AI social proof sensitivity. These informed a survey instrument developed through the qualitative phase and refined against previously validated scales in phase 2, recruiting 442 respondents across 4 Indonesian urban centers. A 14-hypothesis structural model was tested using PLS-SEM (partial least squares-structural equation modeling).

Results: Twelve of 14 hypotheses were supported. The model demonstrated substantial explanatory power for cognitive empowerment (R²=0.661) and parental self-compassion (R²=0.719), while perceived nutrition-oriented feeding practices yielded a more modest value (R²=0.280). This result reflects the multifactorial nature of self-reported nutrition-oriented feeding practices, which is shaped by numerous personal, cultural, and situational factors beyond AI engagement. AI algorithmic trust showed the strongest association with parental self-compassion (β=.479; P<.001), while AI personalization fit perception most strongly predicted cognitive empowerment (β=.275; P<.001). Both intervening variables were significantly associated with nutrition-oriented feeding practices, with cognitive empowerment (β=.232; P=.003) outperforming parental self-compassion (β=.165; P=.009). eHealth self-efficacy moderated the cognitive empowerment pathway (β=.124; P=.02) but not the self-compassion pathway.

Conclusions: The findings suggest that generative AI engagement is associated with perceived nutrition-oriented feeding practices via 2 distinct indirect associations examined in this study: cognitive empowerment and parental self-compassion. These associations were strongest when mothers trusted the AI system and perceived its recommendations as fitting their child and household realities, indicating that the perceived value of AI in toddler feeding may lie less in algorithmic sophistication than in its capacity to build trust, fit context, and support maternal judgment. Because the outcome reflects maternal perception rather than observed feeding behavior or child nutrition outcomes, these findings should be interpreted as associational. They may nonetheless inform future research on AI-assisted feeding support that complements, rather than replaces, professional nutrition counseling.