Brislin's (1970) back-translation procedure is the established standard for translating survey scales in organizational research. However, the procedure has been criticized for relying on the researchers’ subjective evaluation and for its uncritical use. Nowadays, tools based on large language models offer a fast and cost-effective alternative to human translators, with astonishing quality. Across four studies, we examined whether tools based on artificial intelligence (AI) are suitable for the complex task of translating scale items, exemplified by measures from organizational research. After comparing ChatGPT, Google Translate, and DeepL and identifying DeepL as the most suitable AI-based tool (Study 1), we examined the steps of Brislin’s back-translation procedure in detail. Results showed that translations generated by DeepL were comparable to human translations in quality (Study 2), degree of similarity between the original and back-translation (Study 3), and measurement invariance (Study 4), and even superior for some scales. We discuss opportunities and limitations of AI-assisted scale translation and provide specific recommendations for organizational researchers, including the systematic, critical review of translation deviations using the metric for evaluation of translation with explicit ordering (METEOR).
