I believe that when searching for information, the assistant should not be limited to the language of the prompt but should also utilize resources in other languages when relevant. For instance, my native language is Russian, and I find it more convenient to write prompts in Russian. However, I am also aware that Russian-language resources are often of lower quality, and I expect the model to execute queries in multiple languages as appropriate. This is particularly important when searching for political information, where a Russian-language query might retrieve results primarily from sources like rt.ru.
Another example can be seen here: (in Russian) https://kagi.com/assistant/8f823e10-b4df-4b2b-8e1e-3de5c8ac76f0. In this case, the answer is correct, but had the search query been in English, it could have drawn from high-quality English-language medical resources, which are far less abundant in the Russian-speaking web. ChatGPT answer https://chatgpt.com/s/t_696b2d289fe88191aaf293e2470cde80.
And an English prompt concerning a localized event—such as one in Serbia—may lead the assistant to search exclusively in English rather than in both English and Serbian, thereby failing to retrieve the most relevant information.
From my perspective, the output will be more relevant and of higher quality. Its like a Kagi News app that retrieving information from different sources
