This study investigates, if capabilities of text-based AI tools are already able to support the systematic innovation method TRIZ by encompassing the three important elements of analysis, ideation, and evaluation in one go. For this purpose, ChatGPT 3.5—as a prime, free and famous representative—is tested on a sample set of 50 Kangaroo questions. After presenting the overall results of the statistical evaluation the focus is set on detailed aspects and particular issues of some selected questions and answers. Apart from different kinds of answer distributions this also covers exemplary sets of observed inconsistencies and effects such as hallucinations, indicating underlying intrinsic problems. As a key result, the current performance is not satisfying, yet.