Is ChatGPT a change or a problem for translation?

Artificial intelligence has revolutionised the translation industry and ChatGPT is one of the best-known tools in this area.

This advanced OpenAI model translates texts very easily and adapts to the context and style desired by the user.

Unlike other traditional tools, such as Google Translate, it allows for adjustments and specifications in tone, style and vocabulary.

But is this good or bad for professional translators? In this article, we explain how ChatGPT works as a translator, provide some practical tips, show real examples of commands and analyse the impact of AI on translation, including the Portuguese scenario.

Image of a laptop with the ChatGPT logo and a question mark on the screen, accompanied by the phrase ‘Is ChatGPT a change or a problem for translation?’.

What we're going to explore:

ChatGPT as a translator.

ChatGPT is one of the most widely used tools and its performance is increasingly optimised and reliable because it identifies context and expressions. It can provide you with specific instructions to make translations more formal, cultural or technical. What’s more, the model can adjust the style, tone and even adapt the text to different audiences, something that traditional tools don’t do.

ChatGPT has impressive capabilities, but it is no substitute for professional translation. Especially with more technical or complex content, the result still has substantial errors. For this reason, when it comes to cultural nuances or specialised terminology, post-editing of the machine translation by experienced professional translators is crucial. ChatGPT is a tool that helps to translate faster. Translators can then concentrate on proofreading and improving texts.

5 useful tips for using ChatGPT as a translator.

  1. Clearly state the language, language variant, context and type of text.
  2. Use simple language: explain instructions clearly so that there are no misunderstandings.
  3. Always stick to the same style.
  4. Give information about the target audience of the text, the purpose of the content, which channels will be used and the objective.
  5. Adjust the instructions for best results.

The Portuguese ChatGPT - AI models made in Portugal

Image titled “The Portuguese ChatGPT – AI models made in Portugal” highlighting the chatbot Amália, a Portuguese artificial intelligence.

Portugal is investing in specific models for European Portuguese. One of them is the AMÁLIA model, announced by the government and developed by the Centre for Responsible Artificial Intelligence, Nova FCT, Instituto Superior Técnico and Unbabel. This model will include local cultural references and will respect national data protection regulations.

At the same time, national universities are developing open models, such as Albertina and Gervásio, which focus on classification and text generation, respectively. The main aim of these models is to guarantee technological sovereignty and privacy, offering AI solutions that are better adapted to the Portuguese reality.

How AI affects translations

Image with the text ‘How AI affects translations’ on a blue background with graphic elements representing technology and translation.

AI completely changes the way we translate.

  • Better efficiency: first translations are faster, which allows for more detailed revisions.
  • Improved quality: capturing specific aspects and more technical words, allowing translators to concentrate on more complex tasks.
  • Savings: the first automatic translations reduce editing costs.
  • Real-time translation: allows you to speak to the client in several languages at the same time.
  • Market expansion: facilitates entry into international markets.

But there are limitations in specialised, literary texts or those that require precision, such as technical, legal and medical documentation. There are also important concerns about confidentiality and data protection.

Practical examples of prompts in the translation sector.

A good prompt makes a difference to the quality of the translations generated by ChatGPT. A recent study by Dokutech (2025) showed that detailed instructions generate better results in specialised translations. Here are three practical examples:

  1. Technical translation:

    “Translate the technical text below into European Portuguese accurately. Maintain a professional and instructive tone. Steps: a) Read the context and identify key terms; b) Translate using equivalent terms from the sector (see glossary provided); c) Review the syntax to ensure clarity; d) Format the result, mirroring the original structure. Glossary: “maintenance” = “Manutenção”, “hydraulic pump” = “bomba hidráulica”.
  2. Video subtitling:

    “Translate and adapt the following dialogue for subtitles in Portuguese from Portugal. The subtitles should be short (maximum 2 lines) and synchronised with the lines. If the characters are informal, use casual language. Indicate the relevant sounds in parentheses.
  3. Preparation for simultaneous interpreting:

    “Summarise the following conference speech (in EN) and provide a list of the 10 main technical terms with PT translation and approximate pronunciation. Keep the summary true to the main ideas so that an interpreter can understand it.”

How do we evaluate the ChatGPT translation results?

The result of ChatGPT texts is only as good as the instructions and contexts provided. Compared to many traditional tools, the result is far superior, more fluid and coherent.

However, in very specialised content, such as complex technical documentation or literature, it still has important limitations, such as errors in terminology, consistency and cultural nuances. This is why human proofreading is necessary.

FAQs

Illustrative image with the word ‘FAQS’ highlighted on a blue and orange gradient background, representing frequently asked questions about ChatGPT.

Yes, ChatGPT is effective and reliable in most cases, and in certain languages. But for specialised or technical translations, a human review is required to guarantee total accuracy.

Yes, if the prompts are clear and detailed, the quality of the translation improves greatly. However, if the text is too complex or vague, it may not work well.

It depends on the requirement. It is especially useful as a starting point or for non-critical texts. However, for specialised and rigorous content, proofreading by a professional translator is essential.

Yes, Portugal is developing specific models, such as AMÁLIA, and open-source academic initiatives, such as Albertina and Gervásio, to guarantee technological autonomy and data protection.

If prompts are not well structured, are too complex or vague, the quality of the translation suffers. It's important that the instructions are clear, detailed and simple.

Partilhar

Artigos Relacionados