JMIR Publications

ABSTRACT

Technology-mediated services, including chat platforms, social media, mobile applications, and emerging artificial intelligence (AI) tools, are increasingly used to support survivors of intimate partner violence (IPV). These tools can expand access to information and support, particularly for survivors who face barriers to in-person services, such as a partner’s controlling behaviors, geographic distance, transportation, childcare, or concerns about privacy and safety. However, safety in technology-mediated services is not limited to protecting survivors’ privacy and collected data. It also depends on how technologies communicate with survivors. Although risks related to privacy and confidentiality are widely recognized, this Viewpoint draws attention to an underrecognized safety concern: the potential for language used or generated by technology to cause distress, reinforce bias, stereotype, and stigma, or re-traumatize survivors. Language is not neutral. It reflects dominant social norms, power structures, and the perspectives of those with greater access to power, privilege, and resources. As a result, even language that appears respectful or objective may carry bias, stereotypes, victim-blaming narratives, or assumptions that marginalize IPV survivors. Explicitly discriminatory, bigoted, or hateful language may be more readily recognized. More difficult to identify is language that appears neutral but minimizes survivors’ concerns, misinterprets their experiences/thoughts/feelings, implies responsibility for the violence they experienced, or excludes the experiences of male, nonbinary, disabled, racialized, immigrant, or otherwise marginalized survivors. These risks are heightened in digital interactions that rely primarily on written communication because they lack tone, facial expression, body language, and other contextual cues. This concern applies across technology-mediated services but becomes especially urgent with generative AI. Because AI systems are trained on large bodies of historical language data, they may reproduce and amplify entrenched social inequities. Without intentional trauma-informed design and evaluation, AI-generated responses may threaten survivors’ perceived safety and trust in technology, disempower them, and potentially discourage future help-seeking. Drawing on the six principles of a trauma-informed approach, this Viewpoint introduces trauma-informed language as communication that recognizes the widespread impact of trauma, acknowledges that language itself can cause or exacerbate harm, and actively resists re-traumatization through language that promotes safety, trustworthiness and transparency, support, collaboration, empowerment, and attention to cultural, historical, and gendered contexts. This perspective shifts the field from reactive approaches that detect and mitigate harmful outputs after they occur toward proactive prevention. It also reframes language not as a stylistic concern, but as a core design, safety, and equity standard for digital technologies. Future work should develop and test trauma-informed language frameworks, dictionaries, and evaluation criteria across diverse survivor populations and technology contexts to ensure that digital innovation advances not only access, but also safety, dignity, equity, and healing.