Implementation of TB infection prevention and control (IPC) measures in health facilities is frequently inadequate, despite nosocomial TB transmission to patients and health workers causing harm. We aimed to review qualitative evidence of the complexity associated with implementing TB IPC, to help guide the development of TB IPC implementation plans. We undertook a qualitative evidence synthesis of studies that used qualitative methods to explore the experiences of health workers implementing TB IPC in health facilities. We searched eight databases in November 2021, complemented by citation tracking. Two reviewers screened titles and abstracts and reviewed full texts of potentially eligible papers. We used the Critical Appraisals Skills Programme checklist for quality appraisal, thematic synthesis to identify key findings and the GRADE-CERQual method to appraise the certainty of review findings. The review protocol was pre-registered on PROSPERO, ID CRD42020165314. We screened 1062 titles and abstracts and reviewed 102 full texts, with 37 studies included in the synthesis. We developed 10 key findings, five of which we had high confidence in. We describe several components of TB IPC as a complex intervention. Health workers were influenced by their personal occupational TB risk perceptions when deciding whether to implement TB IPC and neglected the contribution of TB IPC to patient safety. Health workers and researchers expressed multiple uncertainties (for example the duration of infectiousness of people with TB), assumptions and misconceptions about what constitutes effective TB IPC, including focussing TB IPC on patients known with TB on treatment who pose a small risk of transmission. Instead, TB IPC resources should target high risk areas for transmission (crowded, poorly ventilated spaces). Furthermore, TB IPC implementation plans should support health workers to translate TB IPC guidelines to local contexts, including how to navigate unintended stigma caused by IPC, and using limited IPC resources effectively.