Background: Three diagnostic criteria have been proposed used for invasive pulmonary aspergillosis (IPA) diagnosis, namely EORTC/MSG criteria, Bulpa criteria and intensive care unit (ICU) criteria. The Bulpa criteria were proposed to diagnose IPA in chronic obstructive pulmonary disease (COPD) patients specially. Our aim is to verify that whether the Bulpa criteria are the most suitable for diagnosing probable IPA in critically ill COPD patients compared with the other two criteria. Methods: We included critically ill COPD patients admitted to the ICU from April 2006 to August 2013. Patients were classified into four populations: population one (n(1) = 59) comprised all included patients; population two (n(2) = 24) comprised patients with positive mycological findings (both positive cultures and positive serologic tests); population three (n(3) = 18) comprised patients with positive lower respiratory tracts (LRTs) isolation; and population four (n(4) = 5) comprised proven IPA patients with histopathology. Patients in four groups were diagnosed as probable IPA using three criteria respectively, and the "diagnostic rate" of each criteria were compared with each other. Then, the reasons for differences in "diagnostic rate" were analyzed in population two. Finally, the modified Bulpa criteria were proposed. Results: Bulpa criteria yielded the highest "diagnostic rate" of probable IPA followed by the ICU criteria, while the EORTC/MSG criteria provided the lowest rates in four populations (the "diagnostic rate" of probable IPA was 33.9%, 16. 9% and 6.8% in population one, p = 0.001; 83.3%, 41.7% and 16.7% in population two, p < 0.001; 100%, 55.6% and 22. 2% in population three, p < 0.001; 100%, 60% and 20% in population four, p = 0.036). The reasons for the highest "diagnostic rate" by Bulpa criteria were its less strict requirements regarding the doses/courses of steroid use and typical computed tomography (CT) findings. Finally, the modified Bulpa criteria for probable IPA were proposed for critically ill COPD patients admitted to ICU, mainly involving revised interpretations of microbiological findings. Conclusions: Among the existing three criteria, the Bulpa criteria are the most suitable for diagnosing probable IPA in critically ill COPD patients admitted to ICU. A modified criteria maybe proposed for better diagnosis, and its clinical validity need to be verified in future studies.