Skip to content

Frequently Asked Questions (FAQ)

Questions and answers

There’s a wide range of language services and an even wider range of language service providers. If you’re not familiar with this industry, you might have trouble telling professional from unprofessional offers and finding suitable candidates for your specific needs.

To make this process easier for you, this page provides answers to some common questions. More items will be added as needed. If your question isn’t answered yet, feel free to drop it into my digital mailbox!

String of lights with several hanging lightbulbs, and the one in front is in focus.

Language Service Providers

Translators work with the written word, whereas interpreters handle spoken text. In addition, we differentiate between simultaneous and consecutive interpreting. With the simultaneous approach, the target version of the text is available almost immediately (example: interpreting services during EU or UN debates), while the consecutive approach means the original speaker and the interpreter take turns (example: interpreting during live events on TV).

Localization specialists are translators, too, but with a focus on software, websites, and related assets. Their goal is to re-create the same user experience in another language. This means all translated elements of the user interface (UI) need to be easy to understand, fit into the available space, and be used consistently in knowledge bases and manuals. Some features or functionality might have to be adjusted to the conventions of the target language as well. Typical examples include dates, currencies, units of measurement, and the reading direction of text.

Subtitlers help you understand foreign-language movies, TV shows, videos on YouTube, reels on TikTok, and so on. They may create captions for people with hearing impairments as well. The process for subtitling and captioning is similar, but there are some differences. For example, subtitles mostly consist of translated dialog, while captions also have to include sounds or similar details that are necessary to understand what’s happening on screen.

General terms such as language consultant or language service provider are often used due to the variety of projects and demands in this line of work. Some tasks may overlap, new technologies require new skills, and more specific terms such as “translator” don’t always reflect the full spectrum of someone’s service offering. This is particularly true for those who provide not just translations but also editing, proofreading, copywriting and related services. If in doubt, clarify with your potential collaborator what exactly you need.

If you need a language service provider, you can check the listings of professional associations for translators and interpreters (for example, the ATA or the German BDÜ), browse directories of translation communities such as Proz.com and TranslatorsCafé, search social media platforms like LinkedIn (relevant hashtags such as #xl8, #1nt or #L10n might help) or run a standard web search for the language combination and subject matter you need, with additional filtering by country/region/city to narrow down the results. (If your search query is too generic, such as “translator English German,” most search engines will clutter the results pages with links to dictionaries, machine translation options, etc., so you need to be more specific.)

It depends. If you only need a handful of languages and don’t have millions of words to translate, professional freelancers will serve you well. Either they cover the respective language combinations themselves or they collaborate with colleagues who can fill the gaps. Even large projects can normally be handled by freelance translators if you notify them early on of any upcoming high-volume tasks or don’t set unreasonably tight deadlines.

Most agencies forward your projects to freelance translators. This means you pay more (to cover the agency’s fee and overhead) but don’t necessarily get higher quality. On the contrary, the quality is often worse. This is particularly true for large agencies that simply send incoming tasks to all translators with a matching language pair and working field in their database and award the job to whoever clicks the “accept task” button first or has the lowest rate per word. And often, those agencies fill their databases by posting generic collaboration requests to translation communities without thoroughly vetting all applicants. So, if your project ends up in the hands of the fastest button clicker or the cheapest provider, you can bet the quality will be low.

Translation agencies are a good option for projects with a very large scope. If you don’t want to deal with the project management and having to find qualified translators for each language pair, an agency has enough resources to assist you.

BUT: Due to the aforementioned quality issues, you should always find out how the agency operates, which criteria are used to assign tasks to freelancers, and what percentage of your payment will actually reach the translators doing the bulk of the work. Trustworthy agencies will tell you the names of their translators and allow direct communication between all involved parties in case of queries or feedback. Shady agencies will promise you the moon while paying their freelancers ultralow rates near or even below the equivalent minimum wage of employees (and without any of the benefits employees would get, of course). This is a recipe for disaster!

Translation Quality

The mother-tongue principle refers to the idea that translators should only translate into their native or first language, not into a foreign or second language. Why? Because it’s easier to express yourself idiomatically and smoothly in your native language than in any foreign one. Therefore, the translations will read better and take less time to create, which ultimately benefits the translation buyer.

However, the mother-tongue principle is not always a hard rule. For rare language pairs or very specific subject matters, it may be difficult to find a qualified native speaker who could handle a complex project until a given deadline. In such cases, correct terminology matters more than flawless style. Besides, native speakers aren’t perfect wordsmiths by birth either – it takes time and practice to learn how to write in an engaging way that keeps the target audience in mind. And sometimes, people’s individual circumstances might even result in them using their native language so little that a second language becomes their first one.

Further reading: “The Mother-Tongue Principle: Hit or Myth?”

You could do that, but the results often disappoint and fall short of requirements. When people talk about “AI” these days, they usually mean so-called Large Language Models (LLMs), which generate plausibly sounding output based on massive amounts of training data and probabilities. But a plausibly sounding response to your question isn’t necessarily correct or suitable for your target audience. LLMs don’t understand texts the way humans do, and they don’t have any of the feelings you’re trying to trigger in your audience with your content.

For fairly simple texts or standardized wording in common language pairs, machines can often generate acceptable translations, but when it comes to more complex texts with carefully chosen words, these systems reveal their shortcomings. Translations tend to be too stiff, literal, and erroneous. Nested sentences, colloquial style, or subtext are typical features of human language that machines fail to render correctly in the target language.

More importantly, the systems referred to as “generative AI” (ChatGPT, Claude, etc.) have devastating consequences for our society and environment. And the initially low subscription prices that aim to turn users into addicts are not economically sustainable either and will rise over time, thereby turning your reliance on those systems into a financial burden as well.

Ultimately, it’s up to you to decide: How important are your texts? What are the potential consequences of mistranslations or data leaks? And do the short-term savings really justify the long-term dependency on dubious “AI” companies?

(For more details about my position regarding “artificial intelligence,” see my AI policy.)

Raw machine translations may be used to speed up translation processes and lower costs. However, this only works well with texts that use relatively simple or generic wording, have many repetitive parts, don’t come with additional requirements such as length/layout restrictions, and will be translated from and into a very common language.

Of course, texts that don’t meet these criteria could be run through MT engines as well, but the necessary effort to review and polish them would be so high that any initial time and cost savings would simply evaporate. And you really should not skip this step. Machine translations are unreliable and should never be published without human review!

Plus, when using “generative AI” such as ChatGPT or Claude, there is zero confidentiality. The companies behind those systems regularly refuse to be held accountable for errors, data leaks or illegal content, and you must assume that any damages will be at your own expense. And it’s your reputation that’ll suffer.

In short: Machine systems require a thorough consideration of risk and reward. If you blindly believe the current “AI” hype or the grandiose promises of big translation agencies, don’t complain later when things go south!

Standards and Tools

Standard lines and standard pages are units of measurement sometimes used in publishing and the translation industry (especially in Germany) to compare the volume of texts regardless of their format. For example, a novel might be available as a compact paperback with a smaller font and as a large hardcover with a bigger font. The total number of pages would be different despite the text being the same. This is why professional translators will ask for more information if you want them to translate your book and only tell them, “my novel has 340 pages.”

To avoid this uncertainty, we can calculate the number of standard lines or standard pages. A standard line has up to 60 characters (including spaces), and a standard page consists of 30 standard lines. Therefore, the maximum number of characters on a standard page would be 60×30=1,800. But a line in a book, magazine or newspaper doesn’t always span the whole page width, so it’s more realistic to say a standard page has about 1,500 characters.

When estimating the cost of translation or editing services, also keep in mind that the total character count calculated by word processing software such as Microsoft Office does not include spaces at the end of lines or at the bottom of pages. To get the true number, you’d have to use a template that sticks to the above definition.

Example: A longer piece written by me in Word had 85 pages, 31,749 words, and 219,257 characters (including spaces). This character count divided by 1,500 results in approximately 146 standard pages. Divided by 1,800, it’s only 122 standard pages. But after using a template and converting my text to the true standard format, it turned out to be 156 standard pages.

Currently, one of the most important standards in the translation industry is ISO 17100:2015 – Requirements for translation services. This standard defines key terms, describes the translation process (all the way from pre-production to post-production), and lists the required qualifications and skills of translators, revisers, and project managers.

Equally important now is ISO 18587:2017 – Post-editing of machine translation output, which focuses on MTPE services. This standard also defines relevant terms, describes the MTPE process, and lists the required qualifications of those offering such services. In particular, it differentiates between “full MTPE” and “light MTPE,” and it clarifies that the aim of the former is “to produce an output which is indistinguishable from human translation output.” Unfortunately, many large translation agencies and some uninformed freelancers often ignore this standard. (For details, see the next item: What is MTPE, and what’s the difference between “light MTPE” and “full MTPE”?)

Older standards that you might still come across include DIN 2345 and EN 15038:2006 for translation services. Both are obsolete now.

MTPE stands for Machine Translation Post-Editing, that is to say, the improvement and polishing of machine-generated translations – a hot topic in the language industry. If this approach is applied in a meaningful way, with suitable texts, it can indeed lower costs and accelerate translation processes.

ISO 18587 differentiates between “light MTPE” and “full MTPE.” The former is only meant to be used for information gisting, and the edited MT output “shall be comprehensible and accurate but need not be stylistically adequate.” Therefore, “light MTPE” should only be used with internal documents but not with content meant for publication! In contrast, the output of a “full MTPE” step should be “indistinguishable from human translation output.”

But there is a major problem: Far too often, big translation agencies do not use MTPE in a meaningful way but rather see it as an excuse to dramatically reduce the fees they pay to their freelance translators. They demand a high-quality translation that’s ready for publication (“full MTPE”) but require translators to review twice the normal volume of text per hour and only pay for superficial editing (“light MTPE”). With predictable consequences: Skilled freelancers stay away from those agencies, and the people who end up editing your texts are forced to rush the job and don’t get paid enough to care.

So, what can you do as a translation buyer? For starters, if a language service provider offers MTPE or “AI” solutions, ask what this really means, how the editing process will be handled (and by whom), and what exactly you’ll be paying for. Ideally, show a sample of the post-edited text to native speakers of the target language and get a quality assessment. Most importantly, never rely on the promises of big translation agencies if you don’t speak the target language yourself!

Project Management

The average throughput when translating continuous text is 500 to 600 words per hour or 3,000 words per day. These are ballpark figures, and the actual numbers may vary depending on language pair and level of difficulty. For example, a complex contract takes longer than a simple user manual. However, sometimes a difficult piece of text might take less time than a seemingly simple one. A typical example for the latter would be individual strings that belong to a software user interface and have been stripped of context, thus forcing the translator to meticulously piece it all back together.

For very urgent projects, the daily doable volume may temporarily be higher, and many translators are also willing to do a nightshift if necessary. But if a single translator promises you a consistent output of 10,000 words per day over a longer period of time, your alarm bells should start ringing! Such a high volume usually results in lower quality.

Keep in mind that the actual turnaround time for your project will also depend on the translator’s schedule. If your selected provider is already busy, even a short text of only 500 words could take longer than a day. Therefore, try to give a heads-up to avoid unwanted delays.

The costs of translation projects depend on various factors, including form, length, and topic or subject matter of your texts, the language combinations and desired turnaround time, the file format and necessary software tools, and any additional requirements such as adherence to a specific glossary or style guide and similar wishes.

Some translators publish standard rates on their websites or in their online profiles. Even if the actual price could end up being a little higher or lower, due to the aforementioned factors, this allows you to estimate the cost of getting your text translated. Usually, translators charge per (source) word, per page or per hour. A project-based flat fee isn’t uncommon either.

By the way: Rates of 1–2 cents per word or less are not sustainable for a professional freelancer. People offering to handle your content for a rate like this are likely to slip you a low-quality machine translation. Even a slightly higher rate will still make it difficult for translators to run a professional business. This is particularly true for Germany and many other EU countries with high costs of living. Translators and agencies working for such low rates have to take shortcuts somewhere without telling you – and by the time you find out, it’ll be too late.

Yes, in many cases this is possible and recommended, especially for high-volume projects. For example, if you’d like to get your whole website and blog translated, you could use a single blog post as a test piece and let your potential language service provider (or multiple providers) translate it to make sure everything will run smoothly and the translation will meet your quality expectations. (Of course, the quality should be assessed by competent speakers of the target language.)

Similarly, many editors of novels and non-fiction books offer to do a sample edit of a few pages. This allows both sides to make sure they’re a good match and to get a better understanding of the required effort.

When it comes to paying for the sample, there are no hard rules. Some service providers will only charge for it if you don’t move forward with the full project. Others will do the sample for free as long as it is not unreasonably long.

Most translations don’t have to be notarized. But if you want to submit translated official documents – such as certificates, deeds or diplomas – to an agency or institution, the translations usually do have to be notarized. That is to say, they need the translator’s stamp or seal of approval to confirm that they’re accurate, complete, and identical with the source version. Not every translator is authorized to do this. (For example, I am not, because notarized documents are rarely required in my working fields.)

You can find qualified professionals offering this service in the Database of Translators and Interpreters of the German state justice administrations.

Professional language service providers ensure the confidentiality of your content. Whether it’s your company’s innovative business strategy, your personal medical history or your latest novel’s thrilling plot – whatever needs to stay confidential will stay confidential.

Unfortunately, there are a lot of unprofessional providers out there, too. Certain low-quality agencies often forward incoming translation requests to everybody with a suitable language combination in their database. There’s no real vetting in advance, and your documents get distributed far and wide. This reckless approach has already resulted in highly confidential corporate documents and personal certificates landing in my mailbox (without any redaction of identifiable information in those files). And who were the senders? Translation agencies I had never worked with before!

Therefore, always do your due diligence and check a translation agency’s references and credentials before handing over confidential data!

And if you consider using machine systems such as ChatGPT, don’t believe for a second that your data will be safe! Even the translation platform DeepL has changed its terms and abandoned its privacy-focused strategy in favor of integrating third-party infrastructure from Amazon.