If you’re looking for information about MTPE (Machine Translation Post-Editing) or anything related to it, let me be upfront: you won’t find a textbook definition here, nor a guide called How to Get Into MTPE and Live Happily Ever After. This isn’t an industry manifesto, and it’s certainly not a nostalgic trip back to the days when translators worked with pencils and people insisted that was real quality.
This is simply a story, told with a touch of humour, based on years of working with real texts, real clients and very real problems. A story about why I don’t offer MTPE services.
MTPE: When the Translator Becomes Optional
Automation in translation has been a reality for quite some time now. Machine translation has swept through the language industry like a tidal wave: machine-generated texts that are all remarkably similar, slightly awkward, predictable, with no real sense of style and good enough.
For procurement departments, it’s almost the perfect solution: cheap, scalable and easy to fit into a spreadsheet.
Then someone had a brilliant idea:
“Let’s add a human at the end to take a quick look.”
And that’s how MTPE entered the picture.
The translator no longer translates. Instead, the job consists of reviewing, approving, fixing a few commas and perhaps correcting the most glaring mistakes before moving on. A bit like quality control in a factory: the role is not to decide how the product is made, but simply to check whether it falls apart when someone taps it.
Naturally, all of this has to be done quickly.
After all, the machine has already done the important part. Human work, as everyone knows, is just a detail.
MTPE: What Are We Actually Talking About?
Technically speaking, MTPE is the process of editing text generated by a machine translation system.
It is not translation, although it is often marketed as exactly that, only “faster and cheaper”.
It’s a bit like claiming that instant coffee is basically the same thing as espresso, just without all the fuss and those expensive La Cimbali machines. You simply stir it well enough.
For very simple content, machine translation can indeed be useful. But as soon as a text becomes even slightly more complex, involving specialist terminology, industry-specific context or wordplay, the machine starts improvising and the process loses the very thing real translation depends on: deliberate language choices.
That’s when the post-editor steps in and wrestles with a text that must be compared against the original, interpreted and checked to see whether it’s fit for purpose.
Preferably at speed.
And this is where theory collides with reality. Usually quite painfully.
Most MTPE projects currently circulating on the market rely on general-purpose machine translation tools. These systems treat a technical manual, a legal contract and a marketing brochure exactly the same way.
The result is predictable: terminology errors, unnatural phrasing and sentences that no experienced translator would ever write voluntarily. This becomes particularly problematic in operating manual translation for export projects, where terminology consistency and technical accuracy directly affect usability and safety.
The same type of problem can be observed in technical documentation projects when information is copied, adapted or only partially updated across different documents. I described one such example in an article about documentation and machine configuration no longer matching, where inconsistencies gradually appeared between manuals, HMI screens and the machine actually installed at the customer’s site.
MTPE Sucks: More Work, Less Money (and Less Enthusiasm)
There is a widespread belief that editing machine translation is easier and faster than translating from scratch.
In reality, the opposite is often true.
Post-editing requires working simultaneously with two texts: the source text and the machine-generated version.
You have to read the original, then read the machine translation, figure out what the algorithm was trying to say, decide what can be salvaged and what should be rewritten entirely, make the necessary changes, check that everything works together and somehow do all of this at a pace that suggests it’s merely a light edit.
The paradox is that this process demands excellent language skills, constant concentration and critical thinking.
At the same time, MTPE rates are significantly lower than translation rates.
The post-editor must constantly decide whether to do the bare minimum or genuinely improve the text, which usually means doing more work for less money.
This is where many post-editors lose their enthusiasm. It’s because the MTPE model itself makes it difficult to reconcile quality, deadlines and remuneration rather than they lack the necessary skills.
No matter how elegantly the marketing materials describe it, MTPE is, at its core, a rather thankless and poorly paid job.
MTPE, Creativity and Quality: I Don’t Really See the Connection
Translation is not merely the transfer of words from one language to another. It also involves making choices, understanding style, adapting content and taking responsibility for the final result.
Machine translation post-editing tends to treat these elements as optional extras. As long as the meaning is more or less clear, the job is considered done. Nobody has time to refine the text, improve it or explore alternative solutions that might make it sound genuinely natural. Speed is the priority. You finish one file as quickly as possible and move on to the next so that the numbers add up at the end of the day.
After several hours of that, any sense of professional fulfilment becomes somewhat theoretical.
Translation Agencies, Clients and the Cult of “As Long as It’s Cheap”
Machine translation post-editing fits perfectly into an industry where quality is difficult to measure but costs are easy to compare.
For intermediaries, it is simply a convenient business model: the element that has the greatest impact on profit margins, namely human work, can be efficiently “optimised.”
Some clients are perfectly happy with this arrangement. It’s cheaper, so it looks like a bargain. At least until someone actually has to read those texts, use them or put their own name under them.
Personally, I prefer working with clients who view language as a communication tool rather than an inconvenient budget line that needs trimming. Clients who understand the difference between a patchwork text and one that has actually been crafted.
Why I Don’t Offer MTPE Services
After many years in the profession, this is not an ideological decision. It is simply the logical conclusion of everything described above.
I do not offer MTPE services because I have no interest in reducing my role to correcting machine-generated texts. I prefer to work with a text from beginning to end and take full responsibility for it, even when that requires more time and effort.
I consciously focus on professional translations from Italian and English into Polish, produced entirely by a human translator.
Yes, that means giving up part of the market. It also means greater professional satisfaction and working with clients who value quality and understand that quality doesn’t happen by accident.
Machine translation and machine translation post-editing are certainly here to stay. In fact, they will continue to develop and will probably roll over large parts of the translation market like a steamroller. But not everything that is fast and cheap is automatically the right solution for every situation.
I chose a different path, one based on experience, expertise and responsibility for the final text.
As is usually the case with such matters, everyone is free to steer their career in whatever direction they consider appropriate.
