BACKGROUND: Intimate partner abuse (including coercive control, physical, sexual, economic, emotional and economic abuse) is common worldwide. Advocacy may help women who are in, or have left, an abusive intimate relationship, to stop or reduce repeat victimisation and overcome consequences of the abuse. Advocacy primarily involves education, safety planning support and increasing access to different services. It may be stand‐alone or part of other services and interventions, and may be provided within healthcare, criminal justice, social, government or specialist domestic violence services. We focus on the abuse of women, as interventions for abused men require different considerations. OBJECTIVES: To assess advocacy interventions for intimate partner abuse in women, in terms of which interventions work for whom, why and in what circumstances. SEARCH METHODS: In January 2019 we searched CENTRAL, MEDLINE, 12 other databases, two trials registers and two relevant websites. The search had three phases: scoping of articles to identify candidate theories; iterative recursive search for studies to explore and fill gaps in these theories; and systematic search for studies to test, confirm or refute our explanatory theory. SELECTION CRITERIA: Empirical studies of any advocacy or multi‐component intervention including advocacy, intended for women aged 15 years and over who were experiencing or had experienced any form of intimate partner abuse, or of advocates delivering such interventions, or experiences of women who were receiving or had received such an intervention. Partner abuse encompasses coercive control in the absence of physical abuse. For theory development, we included studies that did not strictly fit our original criteria but provided information useful for theory development. DATA COLLECTION AND ANALYSIS: Four review authors independently extracted data, with double assessment of 10% of the data, and assessed risk of bias and quality of the evidence. We adopted RAMESES (Realist and meta‐narrative evidence syntheses: evolving standards) standards for reporting results. We applied a realist approach to the analysis. MAIN RESULTS: We included 98 studies (147 articles). There were 88 core studies: 37 focused on advocates (4 survey‐based, 3 instrument development, 30 qualitative focus) and seven on abused women (6 qualitative studies, 1 survey); 44 were experimental intervention studies (some including qualitative evaluations). Ten further studies (3 randomised controlled trials (RCTs), 1 intervention process evaluation, 1 qualitative study, 2 mixed methods studies, 2 surveys of women, and 1 mixed methods study of women and staff) did not fit the original criteria but added useful information, as befitting a realist approach. Two studies are awaiting classification and three are ongoing. Advocacy interventions varied considerably in contact hours, profession delivering and setting. We constructed a conceptual model from six essential principles based on context‐mechanism‐outcome (CMO) patterns. We have moderate and high confidence in evidence for the importance of considering both women's vulnerabilities and intersectionalities and the trade‐offs of abuse‐related decisions in the contexts of individual women's lives. Decisions should consider the risks to the woman's safety from the abuse. Whether actions resulting from advocacy increase or decrease abuse depends on contextual factors (e.g. severity and type of abuse), and the outcomes the particular advocacy intervention is designed to address (e.g. increasing successful court orders versus decreasing depression). We have low confidence in evidence regarding the significance of physical dependencies, being pregnant or having children. There were links between setting (high confidence), and potentially also theoretical underpinnings of interventions, type, duration and intensity of advocacy, advocate discipline and outcomes (moderate and low confidence). A good therapeutic alliance was important (high confidence); this alliance might be improved when advocates are matched with abused women on ethnicity or abuse experience, exercise cultural humility, and remove structural barriers to resource access by marginalised women. We identified significant challenges for advocates in inter‐organisational working, vicarious traumatisation, and lack of clarity on how much support to give a woman (moderate and high confidence). To work effectively, advocates need ongoing training, role clarity, access to resources, and peer and institutional support. Our provisional model highlights the complex way that factors combine and interact for effective advocacy. We confirmed the core ingredients of advocacy according to both women and advocates, supported by studies and theoretical considerations: education and information on abuse; rights and resources; active referral and liaising with other services; risk assessment and safety planning. We were unable to confirm the impact of complexity of the intervention (low confidence). Our low confidence in the evidence was driven mostly by a lack of relevant studies, rather than poor‐quality studies, despite the size of the review. AUTHORS' CONCLUSIONS: Results confirm the core ingredients of advocacy and suggest its use rests on sound theoretical underpinnings. We determined the elements of a good therapeutic alliance and how it might be improved, with a need for particular considerations of the factors affecting marginalised women. Women's goals from advocacy should be considered in the contexts of their personal lives. Women's safety was not necessarily at greatest risk from staying with the abuser. Potentially, if undertaken for long enough, advocacy should benefit an abused woman in terms of at least one outcome providing the goals are matched to each woman's needs. Some outcomes may take months to be determined. Where abuse is severe, some interventions may increase abuse. Advocates have a challenging role and must be supported emotionally, through provision of resources and through professional training, by organisations and peers. Future research should consider the different principles identified in this review, and study outcomes should be considered in relation to the mechanisms and contexts elucidated. More longitudinal evidence is needed. Single‐subject research designs may help determine exactly when effect no longer increases, to determine the duration of longitudinal work, which will likely differ for vulnerable and marginalised women. Further work is needed to ascertain how to tailor advocacy interventions to cultural variations and rural and resource‐poor settings. The methods used in the included studies may, in some cases, limit the applicability and completeness of the data reported. Economic analyses are required to ascertain if resources devoted to advocacy interventions are cost‐effective in healthcare and community settings.