Background: Clean cookstove interventions can theoretically reduce exposure to household air pollution and benefit health, but this requires near-exclusive use of these types of stoves with the simultaneous disuse of traditional stoves. Previous cookstove trials have reported low adoption of new stoves and/or extensive continued traditional stove use. Methods: The Household Air Pollution Intervention Network (HAPIN) trial randomized 3195 pregnant women in Guatemala, India, Peru, and Rwanda to either a liquefied petroleum gas (LPG) stove and fuel intervention (n = 1590) or to a control (n = 1605). The intervention consisted of an LPG stove and two initial cylinders of LPG, free fuel refills delivered to the home, and regular behavioral messaging. We assessed intervention fidelity (delivery of the intervention as intended) and adherence (intervention use) through to the end of gestation, as relevant to the first primary health outcome of the trial: infant birth weight. Fidelity and adherence were evaluated using stove and fuel delivery records, questionnaires, visual observations, and temperature-logging stove use monitors (SUMs). Results: 1585 women received the intervention at a median (interquartile range) of 8.0 (5.0–15.0) days post-randomization and had a gestational age of 17.9 (15.4–20.6) weeks. Over 96% reported cooking exclusively with LPG at two follow-up visits during pregnancy. Less than 4% reported ever running out of LPG. Complete abandonment of traditional stove cooking was observed in over 67% of the intervention households. Of the intervention households, 31.4% removed their traditional stoves upon receipt of the intervention; among those who retained traditional stoves, the majority did not use them: traditional stove use was detected via SUMs on a median (interquartile range) of 0.0% (0.0%, 1.6%) of follow-up days (median follow-up = 134 days). Conclusions: The fidelity of the HAPIN intervention, as measured by stove installation, timely ongoing fuel deliveries, and behavioral reinforcement as needed, was high. Exclusive use of the intervention during pregnancy was also high