<p>In Eastern and Southern Africa (ESA), there is heterogeneity within adolescent mothers’ (AMs’) sexual relationships. However, existing studies focus on single relationship characteristics, even though these characteristics do not operate in isolation. Moreover, no studies have examined multilevel predictors of AMs’ relationships. Identifying and characterizing typologies by multilevel factors can inform targeted interventions to lower HIV risk. Data are from the Population-based HIV impact Assessment (PHIA) Project on AMs (15–19 years) from 9 ESA countries (<i>N</i> = 2,761). We conducted multilevel latent class analysis to identify relationship typologies at the AM and community (PHIA cluster) levels, multinomial logistic regression to characterize typologies by multilevel predictors, and multivariable generalized estimating equations (GEE) log-binomial regression to calculate prevalence ratios of HIV for relationship typologies. There were 3 typologies: <i>married with minimal transactional sex</i> (61%), <i>unmarried peer partnership</i> (30%), and <i>working and high transactional sex</i> (9%). Individual-level (i.e., age, school enrollment, low-wealth household, number of sexual partners) and community-level factors (i.e., urbanicity and proportion of low-wealth households) were associated with typologies. <i>Married with minimal transactional sex</i> was associated with a 51% lower HIV prevalence compared to the <i>unmarried peer partnership</i> (adjusting for multilevel factors). We also identified 2 PHIA cluster-level classes: <i>high age-disparate marriage</i> (49%) and <i>low marriage</i> (51%). Low-wealth communities had higher odds of being <i>high age-disparate marriage</i> communities. Results suggest that AMs’ relationships fall into 3 distinct typologies and 2 community-level classes. Further, HIV prevalence varies across relationship typologies. Utilizing tailored approaches could optimize interventions that reduce AMs’ HIV risk in ESA.</p>