The viability of the human probiotic strains Lactobacillus paracasei NFBC 338 and Bifidobacterium sp. strain UCC 35612 in reconstituted skim milk was assessed by confocal scanning laser microscopy using the LIVE/DEAD BacLight viability stain. The technique was rapid (<30 min) and clearly differentiated live from heat-killed bacteria. The microscopic enumeration of various proportions of viable to heat-killed bacteria was then compared with conventional plating on nutrient agar. Direct microscopic enumeration of bacteria indicated that plate counting led to an underestimation of bacterial numbers, which was most likely related to clumping. Similarly, LIVE/BEAD BacLight staining yielded bacterial counts that were higher than cell numbers obtained by plate counting (CFU) in milk and fermented milk. These results indicate the value of the microscopic approach for rapid viability testing of such probiotic products. In contrast, the numbers obtained by direct microscopic counting for Cheddar cheese and spray-dried probiotic milk powder were lower than those obtained by plate counting. These results highlight the limitations of LIVE/DEAD BacLight staining and the need to optimize the technique for different strain-product combinations. The minimum detection limit for in situ viability staining in conjunction with confocal scanning laser microscopy enumeration was ∼108 bacteria/ml (equivalent to ∼107 CFU/ml), based on Bifidobacterium sp. strain UCC 35612 counts in maximum-recovery diluent.