We were recently asked to compile data on the French education system for a client. This involved, among other things, extracting figures from publications and databases.
ChatGPT provided considerable added value by returning French-language sources even though our initial questions were written in English. For example, the language model identified French databases and provided specific figures extracted from French-language texts. It also explained how to retrieve current values from databases using specific French search terms.
Our conclusion: Language models can be particularly useful for incorporating sources into research even when the researcher does not speak the language in which they are written. Our prompt was formulated in English, yet ChatGPT also returned French-language sources that we had not found in our initial English-language research. Naturally, the results still need to be checked. In this case, we had individual publications translated into German using a translation tool, while a colleague reviewed the French descriptions of the databases to make sure we had selected the correct statistics.
Our recommendation for this type of research: Sources can be restricted by language in the prompt, but this is not always necessary. In our case, we specified only “France” as the region, and ChatGPT nevertheless returned French-language websites.
By Maria Kleiner
![Research in foreign languages: using AI language models to access information from original sources more quickly 1 © [2025] [𝘒𝘐-𝘎𝘦𝘯.] 𝘔𝘢𝘳𝘤𝘦𝘭 𝘖𝘩𝘳𝘦𝘯𝘴𝘤𝘩𝘢𝘭𝘭. 𝘉𝘢𝘴𝘪𝘦𝘳𝘦𝘯𝘥 𝘢𝘶𝘧 𝘞𝘦𝘳𝘬𝘦𝘯 𝘷𝘰𝘯 𝘈𝘯𝘥𝘳𝘦𝘢𝘴 𝘖𝘩𝘳𝘦𝘯𝘴𝘤𝘩𝘢𝘭𝘭.](https://research-ki.de/wp-content/uploads/2026/02/an_owl_and_a_ai_robot_face_looking_at_eachother_ink_dr_97be154f-096b-4b60-a995-3533b440c3e8_1.png)