We numerically investigate the motion of active artificial microswimmers diffusing in a fuel concentration gradient. We observe that, in the steady state, their probability density accumulates in the low-concentration regions, whereas a tagged swimmer drifts with velocity depending in modulus and orientation on how the concentration gradient affects the self-propulsion mechanism. Under most experimentally accessible conditions, the particle drifts toward the high-concentration regions (pseudo-chemotactic drift). A correct interpretation of experimental data must account for such an anti-Fickian behavior.