This study examined the score reliability and validity of observer- (Therapy Process Observational Coding System for Child Psychotherapy—Alliance scale [TPOCS-A]; Vanderbilt Therapeutic Alliance Scale Revised, Short Form [VTAS-R-SF]), therapist- (Therapeutic Alliance Scale for Children Therapist Version [TASC-T]), and youth-rated (Therapeutic Alliance Scale for Children Child Version [TASC-C]) alliance instruments. Youths (N = 50) aged 7–15 (Mage = 10.28 years, SD = 1.84; 88.0% Caucasian; 60.0% male) diagnosed with a principal anxiety disorder received manual-based cognitive–behavioral treatment. Four independent coders, 2 using the TPOCS-A and 2 using the VTAS-R-SF, rated 2 sessions per case from early (Session 3) and late (Sessions 12) treatment. Youth and therapists completed the TASC-C and TASC-T at the end Session 3 and 12. Internal consistency of the alliance instruments was α > .80 and interrater reliability of the observer-rated instruments was ICC(2,2) > .75. The TPOCS-A, VTAS-R-SF, and TASC-T scores showed evidence of convergent validity. Conversely, the TASC-C scores failed to converge with the other instruments in a sample of children (age <11), but did converge in a sample of adolescents (age ≥11). Findings supported the predictive validity of the TASC-T and TASC-C scores. However, whereas the direction of the alliance–outcome association for both observer-rated instruments was in the expected direction for children (negative), the correlations were in the opposite direction for adolescents (positive). Overall, findings support the score reliability of observer- and therapist-report alliance instruments, but questions are raised about the score validity for the observer- and youth-report alliance instruments.