Human Connections in a Digital World

Background/Purpose: Depressive symptoms affect 280 million people worldwide, yet the quality of GenAI translations of depression screeners across languages is unclear. We developed a Translation Validity Index (TVI) to evaluate PHQ‑9 translations produced by ChatGPT, Copilot, and Google Translate in nine languages. Method: Two bilingual evaluators per language (N=18) rated each translation’s cultural appropriateness, grammar, and semantic clarity; TVI scores ≥3 indicated acceptable quality. Results: ChatGPT and Copilot generally met acceptability in high- and medium-resource languages (TVI=3.11–3.66), while Google Translate met acceptability for Ewe (TVI=3.73). Conclusions: TVI provides a structured approach for assessing forward and backward translation quality, but bilingual expert review remains essential when developing accurate mental health measures.