Research has shown that the therapist's contribution to the alliance is more important for the outcome than the patient's contribution (e.g., Baldwin, Wampold, & Imel, 2007); however, knowledge is lacking about which therapist characteristics are relevant for alliance building and development. The objective of this study was to explore the development of the working alliance (using the Working Alliance Inventory), rated by both patients and therapists as a function of therapist in‐session experiences. The therapist experiences were gathered by means of the Development of Psychotherapists Common Core Questionnaire (Orlinsky & Rønnestad, 2005). Data from the Norwegian Multisite Study of the Process and Outcome of Psychotherapy (Havik et al., 1995) were used. Multilevel growth curve analyses of alliance scores from Sessions 3, 12, 20 and 40 showed that the therapist factors predicted working alliance levels or growths differently, depending on whether the alliance was rated by patients or by therapists. For example, it emerged that therapists' negative reactions to patients and their in‐session anxiety affected patient‐rated alliance but not therapist‐rated alliance, whereas therapist experiences of flow (Csikszentmihalyi, 1990) during sessions impacted only the therapist‐rated alliance. The patterns observed in this study imply that therapists should be particularly aware that their negative experiences of therapy are noticed by, and seem to influence, their clients when they evaluate the working alliance through the course of treatment. Copyright © 2014 John Wiley & Sons, Ltd. Key Practitioner Message The findings of this study suggest that the working alliance is influenced by therapists' self‐reported practice experiences, which presumably are communicated through the therapists' in‐session behaviours. The study found a notable divergence between practice experiences that influenced the therapists and those that influenced the patients when evaluating the working alliance. Specifically, practitioners' self‐reported difficulties in practice, such as their negative reactions to patients and their in‐session anxiety, affected patient‐rated alliance but not therapist‐rated alliance, whereas therapist experiences of ‘flow’ during sessions impacted only the therapist‐rated alliance. Practitioners should note that patient alliance ratings were more likely to be influenced by therapists' negative practice experiences than by positive ones. The divergence in the patient and therapist viewpoints has potential implications for therapist training and supervision and everyday self‐reflection.