| In vivo therapist coaching of caregiver-child dyads is a hallmark and primary mechanism of change for Parent-Child Interaction Therapy (PCIT), an evidence-based behavioral parent training program. This study conducted the first systematic review of empirical research of PCIT coaching. Using the Preferred Reporting Items for Systematic reviews and Meta-Analyses, this review assessed how existing PCIT research supports expert coaching guidelines, including (1) study characteristics (e.g., design, phase, coding system); (2) adherence to Funderburk and colleagues’ (2014) Gateways and Stylistic Features; (3) use of, and cross-session changes in, responsive, directive, positive, and critical coaching; and (5) how these variables relate to treatment outcomes. Five databases (CINHAL, ClinicalKey, Google Scholar, PubMed, and PsycInfo) were used to identify eligible studies of original empirical research on PCIT coaching, written in English, that used a standardized observational coding system (Therapist-Parent Interaction Coding System [TPICS] or CDI Observational Assessment of Coaching Highlights Code [COACH] System). Of the 161 screened articles, 11 met full inclusion criteria. Most reviewed studies examined coaching in PCIT’s first phase (91%), cross-sectionally (55%), with the TPICS (91%), English-speaking coaches (91%), clinical cases (82%), PCIT’s standard protocol (63%) from a university clinic (56%), and predominately female caregivers (100%) of standard-age children (78%). Only a few studies examined coaching in PCIT’s second phase, longitudinally, experimentally, with the COACH, and/or other languages. Coded samples generally comported with expert-posited Gateways (e.g., positivity, pacing, specificity), Stylistic Features (e.g., low criticism, low line-feeding/modeling), and other coaching recommendations (e.g., ratios of positive-to-critical and responsive-to-directive coaching, fading, thinning, shaping, and reciprocal determinism), but several had mixed or marginal support, potentially due to differences in coding systems and data reporting. However, Gateways and Stylistic Features–or at least their current operationalizationremain empirically unvalidated, as no study found that they predict salient treatment outcomes. Comparatively, research generally supported other coaching recommendations; most notably responsivity, accuracy, and fading; which typically predict caregiver skill acquisition and treatment completion. Collectively, results clarify the robustness of past PCIT empirical coaching studies and related expert recommendations, highlight areas in need of further research, and clarify best practices in PCIT coaching for therapists, supervisors, and trainers. Keywords: PCIT, in vivo coaching, therapist-caregiver coaching, TPICS, COACH |