JMIR Publications

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 new archetype of caregiver has emerged — the "Alpha Mother," who is highly educated, digitally literate, and increasingly bypassing healthcare professionals in favor of generative AI as a primary toddler nutrition consultant. Yet whether this shift empowers or undermines quality feeding practices remains empirically untested.

Objective:

This study evaluated how AI engagement is associated with perceived toddler feeding practice quality among urban Alpha Mothers in Indonesia. Examining cognitive empowerment and parental self-compassion as mediators, and e-health self-efficacy as a moderator, within a novel AI parenting nutrition alignment approach framework.

Methods:

An exploratory sequential mixed-methods design was employed. Phase 1 comprised semi-structured interviews with 10 Alpha Mothers, with thematic analysis identifying five core constructs: AI algorithmic trust, health information quality, personalization fit perception, information-seeking intensity, and social proof sensitivity. These informed a validated survey in Phase 2, recruiting 442 respondents across four Indonesian urban centers. A 14-hypothesis structural model was tested using PLS-SEM (SmartPLS 4), with predictive validity assessed via CVPAT.

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 toddler feeding practice quality yielded (R2=0.280), reflecting the inherent multifactorial complexity of real-world maternal feeding behaviors. AI algorithmic trust showed the strongest association with parental self-compassion (β=0.479, P<.001), while AI personalization fit perception most strongly predicted cognitive empowerment (β=0.275, P<.001). Both mediators significantly predicted feeding practice quality, with cognitive empowerment (β=0.232, P=.003) outperforming parental self-compassion (β=0.165, P=.009). E-health self-efficacy moderated the cognitive empowerment pathway (β=0.124, P=.017) but not the self-compassion pathway. CVPAT confirmed strong out-of-sample predictive validity (P<.001 and P=.024 vs. respective benchmarks).

Conclusions:

AI functions most effectively in toddler feeding support as a structured decision-support partner rather than an open-ended chatbot. Feeding practice quality shows the strongest association with AI engagement when cognitive empowerment is high and maternal emotional burden is low. These findings support redesigning AI-assisted feeding platforms to incorporate structured intake processes, standardized actionable outputs, and guided prompting templates — implemented as measurable, clinician-collaborated literacy-building interventions that mitigate algorithmic authority and strengthen maternal critical appraisal in real-world toddler feeding contexts.