Generative AI can produce fluent text in seconds, making writing seem easier than ever. But when accuracy, meaning and the reader’s experience matter, good writing still requires considerable thought and effort. AI opens up new possibilities for exploring and improving texts, while shifting much of the work from producing sentences to evaluating alternatives and making decisions. The result may be better writing, even when the process takes longer than expected.
1. Introduction
When we first started using generative AI for writing, we were amazed by how much time it saved us. We could ask a large language model (LLM) to draft an email, improve a paragraph or reformulate a difficult sentence, and within seconds we had something that sounded perfectly reasonable. Writing suddenly seemed much easier.
But our experience has changed. The more we work with LLMs, the more we realise that producing a good text can still take a surprising amount of time. Sometimes we even spend longer on it than we would have before. The interesting thing is that we now spend that time very differently.
2. Not every text needs to be a masterpiece
How much effort we should invest in writing depends largely on what the text is supposed to achieve. A routine email to a conference organiser or an administrative request needs to be clear, polite and accurate. If an LLM can produce a suitable draft in seconds, that’s wonderful. We check it, perhaps change a few words and send it.
But other texts have a different purpose. A scientific paper needs to communicate research findings precisely. A teaching text needs to help readers understand something. An essay published on a website should engage readers, make them think and perhaps encourage them to see a familiar problem differently.
For these texts, fluent sentences are only the beginning. What matters is whether the writing achieves what we intended for the people who will read it.
3. When good-sounding sentences aren’t good enough
One of the most frustrating experiences when writing with AI is receiving a beautifully phrased paragraph that doesn’t quite say what we wanted to say. The sentences flow naturally, the transitions seem elegant and everything sounds convincing. Yet something is wrong.
Perhaps an important distinction has disappeared. Perhaps the emphasis has shifted. Or perhaps the text has become so polished and general that it has lost the particular thought we wanted to communicate.
Then we face an interesting dilemma. Should we accept the attractive wording because it reads so well, or should we change it to recover our original meaning? Sometimes we find ourselves struggling to replace a sentence that sounds excellent with one that expresses our thoughts more accurately.
This has made us much more aware of the difference between writing that sounds good and writing that actually says something.
4. The possibilities have multiplied
The biggest change in our writing process is the number of alternatives we can now explore. Previously, we might have struggled for half an hour with an opening paragraph. Once we had something reasonable, we would probably have kept it. Rewriting an entire introduction or restructuring a longer text simply required too much time.
Today, we can ask an LLM to develop a completely different opening, reorganise the argument or suggest several ways to connect two sections. If the result doesn’t work, we can try again. And again. We can explore approaches that we would never have considered when we had to type every sentence ourselves.
We have even started developing our own style guidelines for different kinds of writing. We specify how we want paragraphs structured, which expressions we prefer, what kind of language to avoid and how the text should address its readers. Our instructions become more detailed as we discover what works.
The results often improve considerably. But each new possibility also requires another decision. Is this version really better? Does it communicate our argument more clearly? Will readers understand it more easily? Does it still sound like us?
5. The work has shifted
Writing with AI has changed where we spend our time. We spend less time struggling to formulate individual sentences and more time thinking about the structure, meaning and overall effect of a text.
This is a remarkable opportunity. We can concentrate on the argument rather than the mechanics of producing words. We can reconsider the order of ideas, explore different explanations and examine whether the text really serves its readers.
But there is a catch. Because revising has become so easy, it is tempting to continue improving the text indefinitely. Another opening might work better. Another structure might be clearer. Perhaps a different conclusion would be more convincing.
At some point, we have to make a decision that no LLM can make for us: this text is good enough for its purpose. Now we stop.
6. Better writing, rather than simply faster writing
We still think generative AI is an extraordinary tool for writing. For routine texts, it can save enormous amounts of time. For more demanding texts, it gives us opportunities to explore, revise and improve our writing in ways that were previously difficult or impractical.
But these opportunities come with responsibility. We need to know what we want to communicate, understand our readers and recognise when an apparently excellent formulation doesn’t express our intended meaning. We also need to decide when further revision is worthwhile and when it is time to finish.
Our experience is that AI hasn’t necessarily made writing quicker. It has made writing different. And when we use it well, it can indeed help us produce a better text, even if getting there takes longer than we originally expected.
How has AI changed your writing process? Do you finish texts faster, or do you find yourself exploring more possibilities and spending more time refining them?
Resources & Further Reading
From the Smart Academics Blog:
- #161: How Should PhD Supervisors Respond When Students Use AI? — Why fluent AI-assisted writing makes the thinking behind a text harder to see.
- #160: Research Communication in the Age of AI: Why Academic Writing Still Matters — Why fluent text production is not the same as meaningful research communication.
Related Learning
Responsible Use of AI for Academic Authors | 2 December 2026
In our Responsible Use of AI for Academic Authors workshop, we explore how researchers can use generative AI thoughtfully and responsibly throughout the academic writing process.
© 2026 Tress Academic
