Research on the identification of poor outcome predictors is crucial for the prevention of therapeutic failure. Previous research suggests that clients' persistent ambivalence is one possible path to unsuccessful psychotherapy. The present study analyses ambivalence—here operationalized as return‐to‐the‐problem markers (RPMs)—in five recovered and five unchanged cases of narrative psychotherapy for major depression. The results suggest that both recovered and unchanged cases presented a similar proportion of RPMs at baseline and a decreasing pattern of these ambivalence markers throughout therapy. However, the decreasing was more accentuated in recovered than in unchanged cases, and at the end of the treatment, the proportion of RPMs of the unchanged cases was significantly higher. The results are discussed in light of previous research on ambivalence in psychotherapy, focusing on the meaning of ambivalence and its clinical implications. Copyright © 2015 John Wiley & Sons, Ltd. Key Practitioner Message Ambivalence towards change, here operationalized as RPMs, seems to be a common process in both recovered and unchanged cases, perhaps signalling the uncertainty and anxiety that change may elicit. Although the number of RPMs decreased in both the recovered and unchanged cases, this reduction was significantly higher in the recovered group. Moreover, at the end of therapy, the recovered group revealed a significantly lower proportion of RPMs than the unchanged group, suggesting that ambivalence resolution (or lack thereof) may play a determining role in the therapy's evolution and outcome. RPMs in later stages of therapy may be operationalized as ‘red flags’ for the therapist to acknowledge the client's stuckness and adapt his or her intervention efforts, turning these instances into developmental opportunities.