25/03/2026

Many companies regularly use AI tools to translate content quickly and cost-effectively. However, this perceived increase in productivity comes at a price as without quality checks, flawed or substandard content can find its way into circulation and cause lasting damage to a brand’s reputation. In this article, you will find out where AI currently provides productive support in translation and where human expertise remains indispensable.

Translations at the touch of a button are nothing new. Ever since the breakthrough of neural machine translation (NMT) in 2016, automatic translation output has been available almost everywhere. Many tools are even integrated into smartphones, browsers or apps.

This development has changed the way companies view language services – with consequences such as:

• Translations are often now produced using tools only
• Declining willingness to pay for professional translations
• The perception of translation as a standardised process that can be automated

With the emergence of generative artificial intelligence (GenAI), the hype has intensified once again. Because AI systems can be specifically influenced by prompts and context, it can quickly seem as though translation can be fully automated.

Machine translations still contain inaccuracies and linguistic errors as well as significant shortcomings. Terminology and context are often not captured accurately, and content is translated word for word. In professional settings, such flaws lead to high levels of post-editing, which quickly negate the anticipated cost savings.

That is why AI translation should be integrated into professional processes and supported by human quality assurance rather than outsourcing translation entirely to an AI tool and simply ‘hoping for the best’.

Possible approaches include, for example:

• Feasibility analysis prior to the project
• Using AI as a resource within translation software (computer-aided translation tools)
• Translation memories (TM) and the use of existing TM matches
• Editing and review by experienced language professionals

This enables companies to produce translations using AI in a targeted manner without compromising on quality or requiring lengthy post-editing processes.

Not everything that is technically possible is necessarily suitable for AI translation. That is why every AI translation project should ideally begin with a feasibility analysis to determine whether machine translation and post-editing are actually suitable for a particular text.

The following questions are important in this regard:

1. Technical: Can the files be processed in a computer-aided translation tool? Is the segmentation appropriate?
2. Linguistic: Is the target language suitable for machine translation with post-editing (MTPE)? Is the source text linguistically correct?
3. Content: Are the text type and subject area suitable for MTPE?
4. Accompanying information: Are there any reference materials, previous translations, style guides or terminology?

Only once these points have been clarified will it be possible to make a realistic assessment of whether and how AI translation can be used effectively.

The requirements for translation processes vary depending on the type of text.

Technical texts are usually written in a neutral style, but contain a great deal of specialised terminology as well as short, context-light fragments such as menu texts or error messages. Particularly in the case of safety-related content, such as operating instructions, a mistranslated word can have serious consequences.

Marketing copy must be easy to read, reflect the brand’s tone, and connect with users on an emotional level. These types of text often also require wordplay or creative phrasing in the target language, which cannot be translated word for word from the original.

Such differences have a significant impact on how well machine translation works and how much post-editing is required. Companies should therefore integrate AI tools into their existing translation processes and work with professional translation resources such as computer-aided translation tools and terminology databases.

But what happens if the necessary processes and terminology for working with AI translations are lacking? Then companies risk the following:

• Content is translated afresh on a regular basis
• The significant amount of post-editing required for checking and final verification offsets cost savings
• Without specified terminology, AI systems do not provide reliable technical terms but instead select the most likely wording
• Terminology corrections can represent up to 45% of the work required for post-editing

This is precisely why it is important not to rely solely on technology but to integrate human expertise as well.

Nowadays, a lot of content can be translated in just a few clicks. We are increasingly hearing people say things like: “We don’t need expensive translators for that. We can do it quickly with ChatGPT or DeepL.”

Unfortunately, companies often cut corners in the wrong places. Contrary to popular belief, systems make mistakes too – and it’s essential that most content translated by AI is checked by humans to ensure it is accurate and of the required quality.

And that brings us straight to the next point: this quality cannot always be assessed straight away. When companies publish content in languages that nobody in-house speaks, it is difficult for them to assess whether an AI translation is actually correct or not. Focusing on (supposedly) good machine translation output for individual languages often leads to the mistaken assumption that all the required languages can be handled effectively using that particular tool.

When properly integrated, AI can speed up translation processes and cut costs. Nevertheless, companies should not dispense with structured workflows, integration in established tools, terminology and human quality assurance: because while technology delivers speed, expertise guarantees reliability. Only by combining professional processes with linguistic and technical expertise can AI translation be put to truly effective use.

Would you like to find out how the team at oneword can make your translation processes more efficient and more reliable by combining AI with human expertise? Then book a no-obligation consultation now.