A Two-Week Plant-Based Challenge: A Randomized Evaluation with Daily Food Logs

A randomized evaluation of a two-week plant-based challenge, conducted by the Alliance Foundation in May and June 2026. Plant-based "challenge" programs, short, time-boxed commitments to reduce or eliminate animal products, are among the most widely used tools for shifting diets, but their evidence base consists almost entirely of uncontrolled surveys of self-selected participants. Of 541 sign-ups, the 458 who completed an enrollment survey were randomized within six strata (gender x consumption group), 80% to attempt the challenge and 20% to eat as usual. Both arms logged the animal products they ate every day for 14 days in free-form text messages or website entries. A two-stage language-model pipeline converted each log to grams of animal protein using a fixed conversion table; 200 daily reports were also blind-coded by hand to validate it (Cohen's kappa 0.92-0.96 on consumption presence, r = 0.88 on daily grams). Thirty days after the study ended, all participants received a follow-up survey. Among the 274 participants who logged consistently enough for analysis, the treatment arm consumed 10.6 g/day less animal protein than control after adjusting for baseline (95% CI [4.3, 17.0], p = .001), a reduction of 31%. At 30 days the arms were no longer statistically distinguishable. Control participants, though asked to eat normally, logged 18% below their own self-reported baseline at the median. This project holds the de-identified participant-level data, the codebook and de-identification record. Raw free-text food logs are not released.