AI Solopreneur TrendDiscover the World's AI Solopreneurs

Oct 27, 2026 · ata-divisions.org

A Marketing Translator Who Uses AI but Refuses to Post-Edit It: Kristal Fellinger

Kristal Fellinger is a German-to-English translator and copywriter who helps German small and medium-sized businesses market themselves in English. In a 2026 article for the American Translators Association's German Language Division, she explained how she uses AI in her work, and why she never simply runs a text through it and edits the result. She also tested Claude and ChatGPT against her own translation. Here is what she hands to AI, and what she keeps for herself.

Few professions have felt the arrival of AI as directly as translation. Machine translation can turn a page of German into English in seconds, and many translators are now asked to "post-edit" machine output instead of translating from scratch.

Kristal Fellinger has taken a different route. She uses AI in her work, but not to produce her translations. She uses it as a sounding board, and keeps the translating itself in her own hands.

This article is for translators and other language professionals, and for anyone whose skill AI appears to threaten. We look at how she uses AI and how she does not, what her own test of Claude and ChatGPT showed, and then, in a separate section, at what her experience still has to do.

Who she is

Kristal Fellinger is a German-to-English translator and copywriter who helps German small and medium-sized businesses reach new customers with English-language marketing. She graduated from the University of Southern California and holds the Chartered Institute of Linguists' Diploma in Translation. In a March 2026 article for the blog of the American Translators Association's German Language Division, she wrote that she has been translating for almost 15 years.

She works under the name Fellinger Communications. The legal notice on her website gives an address in Schörfling am Attersee, Austria, and her website says she has lived in Germany and Austria for over 15 years. The website describes the business in the first person and mentions no employees or partner translators; its translation page says she uses quality assurance tools and a highly qualified proofreader. The site offers translation, copywriting and editing services and an inquiry form for new projects.

Why marketing translation is different

She defines marketing translation as the translation of marketing materials, advertising copy, blog content and corporate communications aimed at customers and other stakeholders. Such texts have to sound natural in the target language and use good copywriting, so translators often adapt the style or the content for the new audience. When the adaptation goes further, it becomes transcreation, a mix of translation and creative copywriting.

How she uses AI

She writes that she does her best work by translating manually and occasionally checking in with an AI tool for ideas when she is stuck on a tricky sentence, word or phrase. She never puts a whole source text into a tool and post-edits the output, because she would miss nuances that come from reading and translating the text herself.

She uses AI for three things:

  • Phrasing suggestions. When a sentence sounds clunky, she pastes in the source sentence, removing identifying information if necessary, asks for five possible translations, and combines elements of her version and the AI's until she is satisfied.
  • Adjusting tone and register. She used to let a translation rest for at least a day or two before refining its tone and style. On a tight deadline, AI gives her some of that distance without the wait.
  • Researching idioms and puns. AI explains unfamiliar idioms, colloquial language and puns in seconds, though she stresses that its explanations must be checked because it can make mistakes.

What her test of Claude and ChatGPT showed

She took a paragraph about a company from one of her actual translation projects, altered it slightly, and gave it to Claude and ChatGPT with a simple "translate this" prompt and a more detailed prompt. She then compared the results with the translation she had produced before ChatGPT existed.

  • Prompts matter. With the simple prompt, she found ChatGPT's translation far more natural than Claude's, though still far from perfect. With the detailed prompt, she preferred several of Claude's phrases.
  • Structure is where AI fell short. The key problem in the paragraph was its last sentence, which, translated directly, jumps from the number of employees to an award and back to the number of employees. In her own translation she restructured it into two sentences. In her test, only ChatGPT with the detailed prompt did something similar.

Her conclusion is that AI cannot be relied on blindly, but that it is useful for brainstorming, rephrasing and researching terms and idioms.

What experience gave her that AI could not

This section is our analysis, not a claim made by Fellinger.

  • Knowing what a sentence should do. Spotting that a sentence that is accurate word for word still does not flow in English, and restructuring it, comes from years of writing for readers.
  • Knowing her audience. Marketing translation is judged by whether it persuades. Adapting tone for a new market is a copywriter's judgment, not a dictionary's.
  • Choosing where AI helps. She uses AI where it widens her options, such as phrasing ideas and idiom research, and keeps it away from the step where nuance is decided.
  • Checking the machine. Her habit of confirming AI's explanations, and of removing identifying details before pasting text, reflects a professional's responsibility to clients.

What you can copy

These are our suggestions, drawn from her story.

  1. Use AI as a sounding board, not a ghostwriter. Ask for several alternatives when you are stuck, then choose and combine them yourself.
  2. Keep the step that defines your value. Identify the part of your work clients actually pay for, and do that part yourself.
  3. Test the tools on your own work. Compare AI output with your best work; the gaps show you where your expertise matters.
  4. Write better prompts. Her test shows that a detailed prompt changes the result.
  5. Protect client information. Remove identifying details before pasting text into an AI tool, and check what it tells you.

Notes on the sources

How she uses AI, her test of Claude and ChatGPT, her years of translating and her degree come from an article Fellinger wrote for the blog of the American Translators Association's German Language Division, published on March 6, 2026, including the author biography at its end. The test results are her own assessment of a single paragraph, and the article does not say which AI tool she uses day to day. Her business name, location, qualifications and services come from her own website. In the ATA article she mentions living in Germany for 20 years; her website says Germany and Austria for over 15 years. Neither source gives client numbers or revenue.

Sources

Related articles