In between my journalistic and academic writing assignments, I also do translation work for a translation company based in Switzerland. I've been working for them for 13 years and although translation work in the last few years has paid very little in comparison with other academic work, if you were quick and efficient, you could still earn quite well.
This has all changed in the last couple of years. Like other translation companies and indeed companies in general, most translation is now done by AI tools. Of course, since these tools are not perfect, the result needs to be proofread or "post edited" by an actual translator. Both customers and translation companies view this "post editing" as "tidying up the text" or "just doing a quick check", because (they argue) it has already been translated. Payment for the "post edit" is thus absolutely minimal, nowhere near what the translator would be earning if they actually translated the text themseves from scratch.
The fact of the matter is that the translator usually has to change the entire text. At the end, they have retranslated the text. In fact, it would have been less work if the text had not been pre-translated by AI at all. And for this re-translation, or actual translation, the translator is paid a fraction of what they would normally earn for a translation.
In short, the "post edit" is a big con. The customers save money while the translator works for peanuts.
Allow me to list some more disadvantages of the post edit for the translator:
1. Researching terminology: Customers should provide a glossary of how they would like terms to be translated. This can be fed to the AI programme and used in the translation. On the whole, however, customers don't provide a glossary (or a comprehensive glossary) and the AI programme uses random terms, which might not be suitable for the customer or the context. This means that the translator has to research every single term anyway - which is the same work as if they did the translation from scratch themselves. The AI programme has been no help here.
2. Consistency: In addition, the AI might translate a term in a certain way in one sentence, but in the next sentence transate the same term in a completely different way. This is a particularly sneaky problem, especially if the term appears at the beginning of the text and then much further down in the text with a different translation, because the translator has to remember that this term was previously translated differently. As both translations might actually be correct, it is not immediately obvious that two different terms have been used and it is up to the translator to try and remember this.
If you are doing a translation from scratch yourself, you will keep your terminology in mind and not use different terms. Consistency is very important in translation. This is different in a post edit situation because you're trying to understand the "mind" of an AI translator which uses no logic whatsoever.
3. Incorrect translations: While many translations produced by the AI tools are quite good, some are utterly wrong. These hit you from left field quite suddenly, as you're not expecting the tool to go off piste quite so drastically. Obviously, these incorrect translations need to be re-done.
4. Style: The AI tool's style is often poor or stilted. This is a very difficult problem as it takes the translator just as long to rewrite these parts of the translation as it would to do the translation from scratch.
5. Hindrance: The AI-translated text is, in short, more of a hindrance than a help in many or even most cases. The translator not only has to check that the translation corresponds in terms of completeness and correctness to the original text, but at the same time bear in mind consistency and style. These are too many elements to take into account simultaneously, so that more than one thorough proofread is always necessary and possibly even three proofreads might be required. The post edit is thus usually more work than an actual translation.
For these reasons, it is very easy to overlook minor mistakes in a post edit. In some cases, if an inordinate amount of changes have been made, to the extent that the translator has spent more time reworking the text than they would have spent translating it themselves, major mistakes might even be overlooked, due to the fact that the translator has been battling simultaneously on all fronts.
At the end of a post edit, I'm usually less satisfied than I am with a translation I've done from scratch (and often frustrated with it).
Conclusion: I therefore appeal to customers - if you really must use an AI tool for a translation, then please respect the translator who is rewriting the text and honour their work with the appropriate fee.
Thank you.

.jpg)