Provenance of the silt on the Chinese Loess Plateau (CLP) is essential for the paleo-environmental interpretation of the thick eolian deposits, but is highly debated. The controversy mainly comes from the multi-interpretation of similar geochemical signatures of potential source areas (PSAs). This work applies the (234U/238U) activity ratio as a new source tracer, because it can distinguish particles with different transporting and recycling histories regardless of petrological origin. The (234U/238U) activity ratio of a fine particle decreases progressively after its production because of the increasing fraction of the 234U precursor ejected out of the particle surface, due to the recoiling effect associated with the α decay of 238U. Distinct spatial patterns of (234U/238U) can be found for the sediments in PSAs and the loess on the CLP. When combined with other constraints, the new findings indicate that the CLP loess can be best explained by the mixing of three end-member dust sources on the northwestern transportation trajectory, namely (1) the Gobi Desert, (2) the Ordos Desert, and (3) the Qilian Mountains. A contribution from the Yellow River is also possible but may not be significant. The identified source partition implies that the eolian silt is mainly produced by processes in the ‘High Asia’ mountains and, partly, by erosion of the exposed clastic rocks. This new constraint on the production and transportation of eolian dust has great implications for the proxy interpretation of loess related to atmospheric circulation, dust accumulation rate, and chemical indexes.