Katharina Happ from GMX and web.de: Journalism from the email provider

Katharina, you are AI manager in the editorial department at GMX and WEB.DE. How did you – originally an online editor – end up in this role, and what fascinates you about the interface between technology and journalism?

Katharina Happ: It actually started with the big ChatGPT moment. When generative AI suddenly became accessible to everyone, I realized very early on how far-reaching its impact would be. At the time, I felt like I was looking out the window and the sky was purple – that’s how dramatic it seemed to me. It was clear to me that this would turn many areas upside down, especially media and text work. And because I found it so exciting, I decided to play a meaningful role in shaping developments in my professional field.

So you went down the rabbit hole?

Happ: Yes, absolutely. At first, I took courses in my spare time, read a lot, and tested things out. At some point, I had gathered so many ideas and so much knowledge that I said to myself: I want to do this professionally. I then accompanied the AI transformation at Funke National Brands Digital, and since April 2025, I have been working as an AI manager at GMX and WEB.DE.

Monitoring current AI developments and developing concrete applications

Your job title is not yet widely known. What does a typical working day look like?

Happ: My working day is very varied. On the one hand, I observe developments in the field of AI, network with other experts, and try out new features. I inform and inspire my colleagues, advise on strategic planning, and consider which use cases could be interesting for us. On the other hand, my focus is on developing specific applications that are tailored to our editorial team and support our everyday work.

What applications are these specifically?

Happ: They are prompts for various tasks: from research aids – such as the precise summarization of extensive scientific studies—to support with headlines to drafts of news articles or press releases that then only need to be fine-tuned. I have now developed almost 40 such highly efficient prompts.

How does such a development process work? Do you work more autonomously or closely with the departments?

Happ: Always in dialogue with the departments. I ask the question at regular intervals: Where do you specifically need support? Sometimes I hear something in an editorial meeting and immediately think: I could build a prompt for that. I then discuss this directly with the relevant colleagues. After that, we clarify the requirements: politics has a more serious tone than entertainment, sports a more engaged tone than knowledge, and then there’s the language of the target groups. I then write the prompt, provide relevant examples and contextual information. I test the draft myself, the team checks it, gives feedback – and even when the prompt is finalised, the optimisation process continues practically endlessly. New study results, AI model developments, and inaccuracies are continuously incorporated so that the prompts are constantly improved.

Have readers ever noticed that AI has helped with a text?

Happ: If an editor’s name appears under the text, it really comes from him or her. AI is used at most in a supporting role, for example in headlines. In the case of AI curation, where the systems are used more extensively, there is a corresponding note. We are not legally required to do so if, as in our case, editorial review and optimization have taken place, but we inform our readers out of conviction.

Journalism at email providers: personalization as a major opportunity

Let’s talk about your business model: GMX and WEB.DE are email providers, but they are also large content platforms. What is behind this combined approach?

Happ: People check their email accounts very frequently. This everyday, familiar environment is ideal for offering high-quality journalistic content. Personalization offers a special opportunity: in addition to our top topics, users see articles that match their interests. This makes our platform a daily companion for many people when it comes to communication and information.

How are these personalized feeds created?

Happ: Personalization is based on two components: the information voluntarily provided during registration and current reading behavior. AI checks for contextual connections and displays suitable articles, all in compliance with GDPR, transparently, and revocable at any time.

How big is your editorial team?

Happ: We have around 70 colleagues in the politics, panorama, entertainment, sports, and knowledge & advice departments who are responsible for the content on GMX and WEB.DE. In addition, there is a community management team that processes around 1,000 pieces of feedback every day. This feedback is structured for internal processes and incorporated into content and further developments.

Your team is based in Munich. Does the location have any special significance?

Happ: Having our editorial team based here is practical for local news. At the same time, we have a national presence; our office in the capital, for example, gives us direct access to federal politics. And as part of 1&1 Mail & Media, we are of course also part of a company that is represented in several cities.

Let’s return to artificial intelligence to conclude. Many journalists are skeptical about AI, whether out of concern for quality or their role. What do you say to colleagues who perceive AI as a threat?

Happ: We shouldn’t be naive: AI will change things, and some activities will shift. There are formats that are primarily about pure information: without classification or complex contexts, with a fairly standardized structure. For content like this, AI offers potential, especially as an efficiency aid, sparring partner, and innovation booster – the latter, for example, in the creation of format versions of self-produced content. But good journalism remains indispensable and cannot be replaced. AI cannot come up with original ideas, provide context based on experience, identify social issues, or uncover new information in interviews. It works on the basis of existing data and probabilities. The original perspective, the connection to the target audience, and personal responsibility must come from humans, and that is precisely what journalism is all about.

Instead of blindly following trends, pause and consider: Which AI trends are truly useful?

There is a preconception that AI can never truly sound like a human being. Do you think that an authentic, individualized, and human-like writing style will be possible in the future?

Happ: Even today, it is possible to a certain extent for AI to imitate the writing style characteristic of a user. To do this, you provide the AI with one or more text samples that reflect the style, tone, and typical phrasing. From these examples, the AI summarizes the typical characteristics in a short, concise “voice paragraph,” which is essentially the voice of the text in condensed form. This paragraph is then integrated into prompts and serves as a guideline so that new texts can be generated in the same style. Nevertheless, you still have to refine it at the end. But as I said, AI cannot offer a writing style colored by socially experienced moments or creative work in the deepest sense of the word.

What is your appeal to the industry regarding the use of AI?

Happ: Things are changing extremely quickly in the field of AI. It’s easy to be tempted to simply follow along and blindly go along with everything. I think it’s important to pause and reflect every now and then. To ask less about how we adapt to AI and more about how AI fits in with us. What content do we want to offer? What values do we represent? And how can AI meaningfully support our work? AI should not determine the identity of a medium, but it can be a valuable tool – if used consciously and knowledgeably.