Background and purpose: Elevation of the chemokine CXCL13 in CSF frequently occurs during active and acute CNS inflammatory processes and presumably is associated with B cell-related immune activation. Elevation levels, however, vary a lot and "leaking" of CXCL13 from blood across dysfunctional brain barriers is a possible source. The aim was to clarify the relation between CXCL13 concentrations in CSF, CXCL13 concentrations in serum and blood-CSF barrier (BCSFB) function for a correct interpretation of the intrathecal origin of CXCL13. Methods: We retrospectively analyzed CXCL13 of banked CSF/serum samples (n = 69) selected from patient records and categorized the CSF CXCL13 elevations as CXCL13 negative (< 30 pg/ml), low (30-100 pg/ml), medium (101-250 pg/ml), or high (> 250 pg/ml). CXCL13 concentrations in CSF and serum and the corresponding CSF/serum CXCL13 quotients (Qcxcl13) were compared to CSF/serum albumin quotients (QAlb) as a measure for BCSFB function. The CXCL13 negative category included two subgroups with normal and dysfunctional BCSFB. Results: Serum CXCL13 concentrations were similar across categories with median levels around 100 pg/ml but differed between individuals (29 to > 505 pg/ml). Despite clear evidence in serum, CXCL13 was detectable only at trace amounts (medians 3.5 and 7.5 pg/ml) in CSF of the two CXCL13 negative subgroups irrespective of a normal or pathological QAlb. Moreover, we found no association between CSF and serum CXCL13 levels or between QAlb and CSF CXCL13 levels in any of the CSF CXCL13-delineated categories. CXCL13 apparently does not "leak" from blood into CSF. This implies an intrathecal origin also for low CSF CXCL13 levels and a caveat for analyzing the Qcxcl13, because higher serum than CSF concentrations arithmetically depress the Qcxcl13 resulting in misleadingly low CSF/serum quotients. Conclusion: We demonstrated that CXCL13 does not cross from blood into CSF, not even during severe BCSFB dysfunction. CSF CXCL13 elevations therefore most likely always are CNS-derived, which highlights their relevance as indicator of inflammatory CNS processes. We recommend data should not be corrected for BCSFB permeability (QAlb) and not to calculate CSF/serum quotients for CXCL13 as these may introduce error.