Skip to content

Translating Business Content with AI: DeepL, ChatGPT or Gemini

Which of DeepL, ChatGPT and Gemini fits translating customer emails, documents and web pages, what to check about your text, and when a human should still read it.

7 min read

Translating Business Content with AI: DeepL, ChatGPT or Gemini

A customer writes to you in Italian. A supplier sends a quote in Japanese. Your own website needs a version for the Vietnamese-speaking families who already ask about you in person. None of these needs a translation agency on day one. All of them are now a few clicks in a tool.

The harder question is which tool, and what it is safe to put in. A translation tool reads every word you give it. For a menu that is nothing. For a contract, a customer complaint or a staff record it is a decision about where that text goes.

This guide compares three tools we review on Find AI Now: DeepL, ChatGPT and Gemini. It draws on our reviews, the vendors' own pages and the Australian privacy regulator's guidance. Sources are at the end.

Quick answer: use DeepL when translation is the whole job, especially for documents, because it is built for it and keeps files in their layout. Use ChatGPT or Gemini when you also want to explain, rewrite or adjust tone in the same conversation. On any free or personal plan, assume your text can be used to improve the service, and keep customer details out. For contracts, safety notices and anything official, have a qualified person check the result.

Two jobs, not one

Business translation splits into two jobs, and the tools split along them.

Converting text. You have a document, an email or a web page and you want the same words in another language. You want it fast, you want the layout kept, and you want the same term translated the same way every time.

Adapting text. You want the message to land with a different audience. Shorter, warmer, more formal, with local phrasing, or explained back to you in English so you know what you are about to send. This is a conversation, not a conversion.

DeepL: the dedicated translator

DeepL is built for the first job. Its translator page says you can drop in PDF, Word and PowerPoint files, and it advertises more than 100 languages. Glossaries, which fix how your own terms are translated, are described as a paid feature on the vendor's pages. The plans page says paid tiers raise the number and size of files you can translate, and that bulk file translation starts at a team tier.

Our review says it is good for translating text and documents, with a writing assistant alongside. If your work is "turn this brochure and these forms into Spanish", it is the shortest path.

The catch, from our review: on the free version the text you put in can be used to improve the service. Paid plans are described as not doing so, and free limits are tight and change. DeepL's own pricing page backs the paid side: it lists no data training for the individual plan and above, and says content is kept only as long as needed to provide the service. It does not make the free position clear, so read its privacy policy before you rely on it.

Skip it if you need to keep confidential text private while on the free plan.

ChatGPT: translate, then talk it through

ChatGPT is not a translator first. It is an all-round assistant that covers writing, images, voice and research, and translation is one thing you can ask of it. Its strength is the second job. You can paste a customer email, ask for a plain English summary, ask for a reply in the customer's language, and then ask it to make that reply less stiff.

Our review says to pick it as one assistant for many jobs. The catch is that the free plan limits access to the strongest models, and the best features sit on higher tiers. It is also one to look at before putting business data in: our review says to skip it if you need business-grade data controls without paying for a team plan.

Note on sources: OpenAI's pricing and privacy pages blocked our fetch, so we have not described its current plan terms here. Check its business privacy page yourself before sending anything sensitive.

Gemini: handy if you live in Google

Gemini suits people who already work in Google apps. It works the same way as ChatGPT for translation: a conversation where you paste text, ask for a version in another language and then adjust it.

The privacy detail matters most here. Google's help page says that with Keep Activity on, your chats are saved and used to develop and improve its services, including the models behind Gemini. A subset of chats is reviewed by human reviewers, and reviewed chats are kept for up to three years. Google's own advice is not to enter confidential information you would not want a reviewer to see. Turning Keep Activity off stops future chats being used to train its models, though Google still keeps chats for 72 hours.

Google's plans page describes separate protections for its business editions: customer data is not used to train Google's models. Our review says to skip Gemini if you need business data protections on a personal account. That is the line to remember: the account type decides the rules, not the tool.

What to check before you paste

One regulator source applies to all three. The Office of the Australian Information Commissioner says organisations should not enter personal information, and particularly sensitive information, into publicly available generative AI tools. Its reasons: it becomes very difficult to track or control how the information is used, and it may be impossible to remove. Where personal information must be used, it says to keep it to the minimum needed.

For a business translating customer correspondence, that gives a short list:

  • Strip names and contact details first. Translate the message, not the person.
  • Use a paid or business plan for anything that has to stay private, and read what it says about training and retention.
  • Turn off history or activity settings on personal accounts where the option exists.
  • Write a rule for staff. Our guide to writing an AI use policy for a small team covers what to put in it.

When a person should still read it

Machine translation is good for understanding and for everyday messages. It is a poor fit where a small error has a cost: contracts, terms, medical or safety instructions, anything addressed to a regulator or a court.

For official use in Australia, check what the receiving body accepts. NAATI is the national standards body for translators and interpreters, and its Certified Translator credential is the one it currently offers. Which documents need a certified translation differs by agency, so ask the agency, not the tool.

For marketing copy, have a native speaker read it before it goes live. The meaning can be right and the tone still wrong. A good habit is to translate back into English with a second tool and compare. If the two versions disagree, find out why before publishing.

Where to start

Sources

Share this article

All articles