The cross-bleaching behaviour of automated Risø TL/OSL (DA-12, DA-15, DA-20) luminescence readers is investigated. By design, up to 24 or 48 aliquots can be stored on a carousel in a single measurement chamber. Due to this construction, irradiation or illumination on one sample may affect the adjacent position resulting in systematic errors. Previously reported for blue LEDs, such cross-talk (cross-bleaching/illumination) has never been quantified explicitly for the infrared (IR) LEDs, although they are intensively used in IRSL measurements of e.g. feldspar and polymineral samples. In IRSL measurements of feldspar or polymineral samples it is important to keep the time constant between the (midpoint of the) irradiation and the subsequent read out to avoid the malign effects of anomalous fading in laboratory constructed dose response curves. This may be achieved by running all measurements for equivalent dose estimation on a single sample before moving to a subsequent sample (e.g. by using the “run 1 at a time” option in the Risø sequence editor). However, if the measurement sequence is not designed carefully, then using this option may result in a significant depletion of the natural signal on subsequent samples. Here we investigate the size of this reduction due to cross-bleaching from the IR diodes and quantify the cross-bleaching for 10 different Risø TL/OSL readers produced between 1994 and 2011. We find that cross-bleaching from the IR diodes is worse than from the blue diodes. Using the “run 1 at the time” option can result in significant dose underestimation (1) if the sequence is not split into different sets, or (2) if samples are not placed on every 2nd position. In addition, a newly designed flange for the optical unit of the TL/OSL reader is presented which appears to reduce cross-bleaching significantly.