Abstract / Summary
Abstract Background Digital health tools can widen the disparities they aim to close when benefiting from them requires unequally distributed resources. This study tested whether socioeconomic status (SES) structures the adoption of a live sexual and reproductive health (SRH) chatbot among digitally included young adults. Methods A national sample of 1200 U.S. young adults (18 to 32) completed a guided interaction with Roo, Planned Parenthood Federation of America’s (PPFA) free SRH chatbot, and a survey measuring nine adoption constructs. Analyses across near-equal education bands ( n = 406, 400, and 390) comprised composite comparisons, Spearman correlations, measurement invariance testing with permutation structural comparisons, a bootstrapped SES resource-chain model, chi-square tests of stated needs, and qualitative content analysis of 110 open-ended responses, and demographic contrasts (sexual orientation, ethnicity, race, gender). Results Gradients in readiness resources were small (significant Cohen’s d between − .15 and − .24 for effort expectancy, self-efficacy, eHealth literacy, and social influence) and the gradient in adoption intention was absent (d = − .09); all Spearman correlations of education and income with the nine constructs fell at or below .09. Measurement was invariant across the lowest and highest bands, and none of eleven permutation-tested structural path differences survived false-discovery correction, although composite variances for literacy, effort expectancy, and anthropomorphism were larger in the low band. A resource-chain model located the entire small total SES effect on intention (β = .064) in an indirect pathway through chatbot-specific eHealth literacy and self-efficacy (indirect β = .068; direct β = −.004). No demographic group showed an adoption penalty: sexual-minority participants rated it easier and more humanlike, and racially minoritized participants reported higher intention alongside higher perceived barriers. Stated needs differed in content, not access: privacy and accuracy concerns were universal, device availability was flat, lower-education participants asked for speed and visual clarity, and higher-education participants for verified accuracy and clinical integration. Open-ended accounts centered on response quality, human-care preference, and privacy assurance; none named device, connectivity, or cost as a barrier. Conclusions Conditional on digital inclusion, the chatbot’s front door opened equally. The remaining inequity sits upstream at first-level access; the one active conduit, chatbot-specific eHealth literacy, is teachable.