Background and purpose: Recent findings document a blunted humoral response to SARS-CoV-2 vaccination in patients on anti-CD20 treatment. Although most patients develop a cellular response, it is still important to identify predictors of seroconversion to optimize vaccine responses. Methods: We determined antibody responses after SARS-CoV-2 vaccination in a real-world cohort of multiple sclerosis patients (n = 94) treated with anti-CD20, mainly rituximab, with variable treatment duration (median = 2.9, range = 0.4–9.6 years) and time from last anti-CD20 infusion to vaccination (median = 190, range = 60–1032 days). Results: We find that presence of B cells and/or rituximab in blood predict seroconversion better than time since last infusion. Using multiple logistic regression, presence of >0.5% B cells increased probability of seroconversion with an odds ratio (OR) of 5.0 (95% confidence interval [CI] = 1.0–28.1, p = 0.055), whereas the corresponding OR for ≥6 months since last infusion was 1.45 (95% CI = 0.20–10.15, p = 0.705). In contrast, detectable rituximab levels were negatively associated with seroconversion (OR = 0.05, 95% CI = 0.002–0.392, p = 0.012). Furthermore, naïve and memory IgG+ B cells correlated with antibody levels. Although retreatment with rituximab at 4 weeks or more after booster depleted spike-specific B cells, it did not noticeably affect the rate of decline in antibody titers. Interferon-γ and/or interleukin-13 T-cell responses to the spike S1 domain were observed in most patients, but with no correlation to spike antibody levels. Conclusions: These findings are relevant for providing individualized guidance to patients and planning of vaccination schemes, in turn optimizing benefit–risk with anti-CD20.