What is generative AI, and what can you do with it?

#2.1How do you recognize generative AI?

Generative AI is now woven into your daily life, even if you are not consciously aware of it. Maybe you use Google Docs and a suggestion for your next sentence appears automatically, or you see an automatic completion for a word you are typing in WhatsApp. But generative AI goes further: it generates entirely new text, images, code, or sound based on a prompt you enter.

A large part of this technology is based on so-called large language models (LLMs). These models are trained on large datasets from the internet, publishers, and more. Based on that, they learn how people communicate. In doing so, generative AI does not use literal copies of the training data, but learns patterns and relationships through mathematical models. Like a writer inspired by books, AI generates new content based on what it has 'read'. They generate text by repeatedly predicting as well as possible which word is most likely to follow the previous word. The result often feels surprisingly fluent, as if you are talking to a person. But it is nothing more than statistics at scale (Brown, 2020). You will learn more about this in chapters 5 and 6.

You can recognize generative AI by output that seems original, yet is always based on patterns from training data. A text generated by ChatGPT is never copied directly from a source on the internet, but is composed based on what the model has 'learned' people are likely to write.

#2.2What can you do with it? Applications of generative AI

Keeping in mind the rules and guidelines of your faculty or course, generative AI can offer you concrete support in your studies and in many other situations. During the writing process for an assignment, for example, you can use AI for brainstorming, structuring texts, or checking language use. When creating visual presentations, you can use AI to generate images, charts, or even illustrations. In coding environments, you can use AI to generate sample code, entire websites, or fix bugs. AI can help you optimize your study process in various ways, such as:

• Automatically summarizing and organizing notes from lectures or conversations for a better overview.
• Turning your books or online texts into podcasts so you can learn on the go.
• Asking focused and substantive questions about your course material.
• Serving as a personal study coach that motivates you, helps you plan, and guides your learning process.

You can also use generative AI for language support, for example by practicing your writing skills with alternative phrasing or asking for targeted feedback on your draft text. Some students even use AI as a speaking partner when practicing foreign languages or role plays, or get help preparing presentations by generating scripts.

AI can also be used as a feedback partner in fields where language is crucial, such as history, law, or communication. You can have concepts explained in simpler language or ask for alternative wording. In more creative contexts, such as art or design, you can use AI to turn rough ideas into visual or auditory versions.

#2.3Limitations and misconceptions

A common misconception is that generative AI actually thinks or understands what you ask. In reality, it does nothing more than calculate which word is most likely to come next. The model does not understand irony, emotion, or context in a human way (at least not yet). This sometimes leads to hallucinations: convincing but factually incorrect text. So it is always important to check the information and facts an AI creates against external sources. And even if you ask the generative AI for sources, make sure those sources actually exist.

An important risk in using AI is that generative AI reproduces social or cultural biases. For example, when a model is trained mainly on English-language, Western data, it may exclude worldviews and perspectives or be stereotypical. Generative AI not only reproduces existing biases, but often does so on the basis of training data that partly comes from unreliable or even problematic sources. For instance, a widely used dataset (mC4) for training ChatGPT also included websites with conspiracy theories or extremist content (Hofman & Veerbeek, 2023). This creates a risk that worldviews are distorted or excluded in the output. You can read more about the origin, quality, and legality of these data in chapter 5.

Another misconception is that generative AI replaces human creativity. In reality, AI often works better as a human assistant. It is a co-creative tool: you lead, the model follows. For example, generative AI only starts to produce truly creative ideas when you ask it to connect concepts that seem unrelated. You give the instructions and must be able to judge whether something is useful, correct, and ethically sound. In addition, working with generative AI is often an iterative process, in which you are not done after a single prompt. You get the best results when you build on the first result with your own creativity by giving more precise follow-up instructions.

Finally, generative AI can affect your learning process. Using generative AI critically begins with self-knowledge: know when you need AI, and when you do not. Ask yourself questions such as: do I really understand this assignment, or am I letting the AI's suggestions guide me too quickly? Am I using AI to learn, or to avoid work I could actually learn from?We will go into this in more detail later.

#2.4Trust in your work

What about using generative AI to produce texts or other work? Even if you say that you use it and explain how (in other words, you are transparent about it), doubts can still arise about you as a person. Dennis Vaendel (2025) writes in Maastricht University's newspaper that cover letters that are visibly written with the help of generative AI immediately raise doubts about the candidate's suitability. In an article in De Correspondent, Michel de Hoog (2025) paints a more complex picture based on research. It suggests that people who admit to using AI for presentations, reports, or emails are seen as less capable and trustworthy, even if their work may actually be better. This creates a paradox in which honesty about AI use - normally good for trust - backfires, and many workers therefore hide their AI use as ‘secret cyborgs.’ But anyone caught doing this risks even greater reputational damage, as shown by examples of lawyers and writers whose careers were harmed. The question remains whether this distrust is temporary and will fade as AI becomes more commonplace, or whether doubts about authenticity and effort will persist. What do you think?