Using GenAI for Writing

Writing is arguably the most important skill in academia. Of course, it goes without saying that you should know how to do research. However, even if your science is solid but your paper is poorly written, it would never be accepted in a top journal. Unfortunately, the reverse isn‘t always true and I have seen polished papers in top journals containing fundamental analytical flaws.

Another important aspect of academia is that it’s a reputation-based institution. The quality of your written work directly influences how colleagues perceive your intellectual capabilities and professional standards. Your social connections are essential for a successful career, and these connections depend on your reputation.

Given the importance of writing skills and reputation, you should be extremely careful when it comes to writing. This caution goes beyond merely following institutional policies or journal guidelines. Even if a certain venue allows using Generative AI in writing, but your colleagues feel that your language is too “ChatGPTish” and they dislike it, this could cause serious reputational damage. It wouldn‘t matter if it was allowed or not. You don’t want to develop a reputation as a lazy, sloppy writer.

At the same time, given the importance of writing and that so many researchers are non-native speakers, the academic publishing system inherently generates significant inequalities. Non-native speakers find it harder to get their work published, regardless of the quality of research. GenAI could potentially serve as an equaliser helping people who are not native English speakers to polish their writing. In fact, Springer Nature explicitly allows for using AI-assisted copy editing of manuscripts without requiring disclosureEditorial Policies: AI Authorship .

With this in mind, I would generally recommend using GenAI to improve your writing rather than trying to save time on writing. Personally, I often use Generative AI for writing but I don‘t think I spend less time on it. I still work a lot on my texts drafting, revising, and editing them. What is different is that the final outcome is probably better thanks to access to Generative AI. Since writing processes are highly individual, there cannot be universal advice for integrating GenAI into your workflow. However, I’ll share my approach, which you might find useful as a starting point for developing your own.

Writing advice

My general advice is to approach GenAI as a developmental tool that improves your writing skills over time, rather than as a substitute for your own writing process. A useful analogy here is the concept from educational psychology known as the zone of proximal developmentThe Zone of Proximal Development (ZPD) , originally developed by Vygotsky. The idea is that there are things a learner can do unaided, in which case there is no learning as the person has already mastered the necessary skills. There are also things that learners cannot do even with assistance, and again no learning occurs here because the material exceeds the learner‘s current developmental capacity. Between these two zones lies the zone of proximal development—where a person can accomplish something with help from others, and this is where actual learning takes place.

In the context of AI, this means you should not move too far from what you’re currently capable of when relying on AI assistance. For example, if AI suggests a word but you‘re uncertain about its precise meaning, I would strongly advise against using it. This would be inauthentic, and there’s a risk that the phrasing would not properly capture your thoughts. However, if AI suggests a word that exists in your passive vocabulary but not your active one—something you understand well but wouldn‘t typically use yourself—this could be an actual improvement for your text. Moreover, it could help move this word from your passive to active vocabulary, thus developing your writing skills over time.

My own process always begins with a draft that’s completely my own. The only thing that changed since the emergence of GenAI tools is that I now usually start by speaking rather than writing. And then this audio is automatically transcribed, cleaned, and split into paragraphs.

Once I have this first draft, I start refining it. When I feel it‘s close to what I want to say, I ask AI to polish the text, always instructing it to stay close to the original, not to introduce new ideas, and to maintain my style. If I like certain suggestions, I incorporate them back into my text and continue editing. If I’ve made substantial changes and become uncertain about my English again, I might ask GenAI to proofread it once more.

This approach still involves considerable work with the text but I find it has helped me a lot. First, it improved the English in my writing. Even more importantly, it removed my writer‘s block. Previously, I could spend a lot of time struggling with just one sentence, which often demotivated me from working at all. Now, starting is much easier—I simply talk, there is no even need to type anything. And I’m not afraid of encountering complicated sentence as I can always ask AI for advice: “What’s the best way to express this in English?” and build upon the suggestions.

This approach has changed my writing process while ensuring that my voice and ideas remain authentically mine.