This paper aims to examine the characteristics of Cicero’s probabilism in his philosophical dialogues, especially in Academica. His philosophical dialogues express a moderate sceptical attitude which allows to endorse persuasive views on a provisional basis. He suggests that his inquiry would reveal the view which is ‘most likely to be true’. At first sight, when he makes a choice between available theories in many places of his philosophical dialogues, he seems to do it on an arbitrary basis without giving any reasons. However, although he does not explain this characteristic process of investigating persuasive views in any detail, I suggest that the discussions in Academica would show characteristic features of Cicero’s moderate sceptical approach to theoretical discussion. This method also has a close affinity with Carneades’ procedure to test persuasiveness of an impression. Thus the ready acceptability of any theory seems to be taken as an initial sign of its persuasiveness. Moreover, some persuasive views exert more convincing power on us than others, when they have gone through further tests of consistency. Even when one philosophical view is impeded by another conflicting view, the one is not removed entirely from investigation. Indeed, there is no way to confirm that any view is decisively false, because there is no way to tell the veracity between them. Thus, even an refuted view can re-enter the inquiry at any time. The only view which Cicero can present more or less certainly as unimpeded would be the case of akatalēpsia. In principle, this is also open to impediment by other views. However, the idea of an ‘unimpeded’ view does not refer to the final result of investigation. Therefore, by ‘following the unimpeded persuasive view’, Cicero must mean ‘follow what is persuasive as long as it is unimpeded’. This is a state of continuous investigation, not an outcome of the investigation.