Is ChatGPT good at translation? How to use the chatbot to translate languages

AI chatbots like ChatGPT have changed the way many of us work, thanks to their ability to quickly understand huge amounts of information. The chatbot’s versatility extends to a staggering number of tasks, ranging from writing essays to even software development. But most of us have only ever interacted with the chatbot in our primary language, so how well can it understand other languages? Surprisingly well! In fact, you can also use ChatGPT for translation between languages.

Is ChatGPT good at translation?

Yes, ChatGPT does a great job at translation even if it’s not officially meant to perform as a translator. The chatbot was trained on a variety of text sources, so it can understand and respond in most major languages. You may wonder why anyone would use ChatGPT as a translation tool instead of a dedicated app like Google Translate. The answer’s simple: ChatGPT was designed as a chatbot first and foremost, so it excels at conversational tasks. For a real-world example, take a look at the screenshot below.

Here, I simply asked ChatGPT to translate a simple greeting from English to Hindi. The chatbot went above and beyond, also offering advice for informal conversations. Since I speak both languages, I can confidently say that this is genuinely helpful advice. Using one translation versus the other could completely change the tone of a conversation. I’ve also written extensively about my experience using ChatGPT for translation in various Asian countries and why I usually prefer it vs Google Translate the vast majority of time.

As the above screenshot shows, ChatGPT also explains when I might want to opt for the second translation. Such features are completely absent from apps like Google Translate, making ChatGPT a good choice. I also tend to translate text more often now that the ChatGPT mobile app is available. It’s more convenient than opening a new browser tab and logging in each time.

ChatGPT vs Google Translate: Which is better?

Large language models have been trained on billions of text samples, across dozens of languages, so it’s not surprising that they do a good job of translation. In fact, Google’s own technical report of its PaLM 2 model admitted this fact. The model scored better than Google Translate when tasked with translating popular languages like Portuguese and Chinese.

A group of researchers also arrived at the same conclusion; ChatGPT’s translation can match or even surpass the quality of Google Translate. However, the chatbot didn’t perform as well when asked to translate technical, complex, or formal sentences. In such cases, you may prefer Google Translate’s generic approach to translation.

Broadly speaking, ChatGPT and other AI chatbots like it are certainly capable of translating between languages. And with future models like GPT-5 and Google’s Gemini on the horizon, things will only improve from here on out.

How to use ChatGPT for translation

In order to use ChatGPT for translation, use a prompt like “Translate [text] into Chinese”. If you have a longer piece of text, like a medical report or job description, use a prompt like “Translate the following into Spanish: [text]”.

Apart from those basic prompts, here are some more tips to help you get the most out of using ChatGPT for translation:

• ChatGPT can do a good job of simplifying its translation if you’re not very familiar with the target language. Use a prompt like “Translate [text] into Spanish for someone with basic language proficiency”.
• If you don’t understand the meaning of a particular translated word, copy and paste it into a new prompt and ask ChatGPT to clarify its meaning. For example, “I don’t understand why you used the word những, what does it mean?”
• You can specify your audience and gender before you ask for a translation. The latter can come in especially handy if your target language uses gendered words (like French or Hindi).

In case you aren’t satisfied with the translation quality, I’d recommend giving rival chatbots like Google Bard a try. A different large language model may have more training data to accurately translate less-widely spoken languages.

FAQs