BACKGROUND: Community health workers (CHWs) in low- and middle-income countries are key to increasing coverage of maternal and newborn interventions through home visits to counsel families about healthy behaviours. Household surveys enable tracking the progress of CHW programmes but recent evidence questions the accuracy of maternal reports. We measured the validity of women's responses about the content of care they received during CHW home visits and examined whether the accuracy of women's responses was affected by CHW counselling skills. METHODS: We conducted a criterion validity study in 2019, in Gombe State-Nigeria, and collected data from 362 pregnant women. During accompanied CHW home visits the content of CHW care and the presence or absence of 18 positive counselling skills were observed and documented by a researcher. In a follow-up interview three months later, the same women were asked about the care received during the CHW home visit. Women's reports were compared with observation data and the sensitivity, specificity, and area under receiver curve (AUC) calculated. We performed a covariate validity analysis that adjusted for a counselling skill score to assess the variation in accuracy of women's reports with CHW counselling skills. RESULTS: Ten indicators were included in the validity analysis. Women consistently overestimated the content of care CHWs provided and no indicator met the condition for individual-level accuracy set at AUC ≥ 0.6. The CHW counselling skill score ranged from 9-18 points from a possible 18, with a mean of 14.3; checking on client history or concerns were the most frequently missed item. There was evidence that unmarried women and the relatively most poor women received less skilled counselling than other women (mean counselling scores of 13.2 and 13.7 respectively). There was no consistent evidence of an association between higher counselling skill scores and better accuracy of women's reports. CONCLUSIONS: The validity of women's responses about CHW care content was poor and consistently overestimated coverage. We discuss several challenges in applying criterion validity study methods to examine measures of community-based care and make only cautious interpretation of the findings that may be relevant to other researchers interested in developing similar studies.