The most important part of writing is still editing

A few weeks ago, I got through to the end of the first draft of my next book. There are lots of bits that need revisiting, but now I can see the full shape of it. I wrote it in the early hours of the morning, between April and October. While I did have a visiting fellowship in Sweden, I had no writing retreats, and no time that was solely blocked out for this project. I wrote it on top of everything else — some upper level teaching, parenting and household management, a couple of research projects, visiting fellow tasks such as being a doctoral opponent/discussant in Sweden. I am tempted to congratulate myself — and indeed, did go out for glass of champagne with my mum, sister and niece to celebrate on the day I got to the end. But I know the hard work that is still ahead. Even in the age of AI, the most important part of writing is still editing.

Editing in the age of AI

When large language models first became available to the public in 2023, I very quickly started playing around with them to figure out what parts of my job they could help me with. One useful part is for editing and summarising text that I have written. On one memorable occasion, I had to edit an abstract for a piece about to be published (the journal was one of the few still using email for submissions rather than an online system and the response piece had managed to get all the way through review with an abstract that was too long). By this point, I had started to dislike the generic voice of ChatGPT, so I prompted it “Retaining my voice and writing style, reduce this abstract by 50 words”. Success! After a few tweaks to make sure the key ideas were still there, I submitted it. I think it probably saved me a few hours of work.

Another way I used LLMs in the early days was to generate bullet point summaries of chapter sections as required in a textbook I was editing. The authors themselves were supposed to submit these, but after repeated requests, I still had a number of chapters that did not have them. For these, I cut and paste chapter sections into ChatGPT and asked for bullet point summaries suitable for undergraduate students. I then sent these back to the wayward authors and asked them for approval or edits. That saved me a huge amount of time at a critical point in the editing process.

But from 2025, I have started to become concerned about using AI to edit. I have noticed that when you put something in that is actually quite good and very clear, it still makes suggestions, and the suggestions invariably — because it is based on majoritarian statistics — make the writing and the ideas more homogeneous. It consistently oversteps from proofreading to line editing to content editing. As I have watched students become more and more reliant on it, I have seen that rather than helping them build capacity, it is undermining their confidence. The entire model is set up to encourage ongoing engagement, so it is hardly surprising. If you ask for proof reading, you will get content editing offered. If I ask for fact checking, I will be offered editing. Not just offered — but paragraphs of suggestions that go far beyond what I asked it to do (for the record, I now use CoPilot, which is the more secure service offered in our workplace). As an editor and an author, I have also seen peer reviews that seem to be generated by AI making overreaching suggestions that do not seem relevant to the quality of the research in the article. Obviously, we can choose whether we use the suggested content or not. But I think its smooth suggestions can be both enticing and destabilising in much the same way junk food is.

The second issue for me is that not only is writing the actual process of thinking, but editing is also the thinking. When we edit, we are making decisions about what is important, what we are trying to say, and how we will say it. When I use a particular word, I am considering the implications of that word and what I really mean by it. When I change that word or the words around it, I am starting to understand what I actually think. We cannot outsource our thinking to a large language model based primarily on statistics. While there is some evidence that AI helps a single individual come up with more ideas and perspectives overall than they would on their own, there is also evidence that as more people use AI the ideas across cohorts end up being relatively homogenised (see this systematic review by de Rooij and Biskaer, 2026). And for myself, this is not what I’m wanting to see as I think our global “Business as Usual” is deeply problematic and we need to rethink it, drastically.

“not only is writing the actual process of thinking, but editing is also the thinking”

This is not to say that all editing must be done alone, or that if we cannot write a book all by ourselves, we do not know how to think. I am fully committed to collective thinking and writing, and also editing. I understand the arguments that you can train an AI agent to do particular things, to be a particular kind of AI in dialogue with different rationalities (see the work by the Meta-Relationality Institute, and James Bridle’s book Ways of Being for examples). But I am also deeply interested in the generalised deskilling we see across the world and invested in retaining the embodied skills of writing and editing for future generations.

So for this book, I choose to work in partnership with human editors for the editorial process — I have a book mentor (the wonderful Professor Tara Brabazon) who has been very generous doing line edits and asking developmental questions all the way through the writing process. And I am now managing my own editing process after having this first draft of the book. At some point, I will engage a copy editor (usually the talented Mona-Lynn Courteau). This is not to say that AI has not been used at all in the book writing process — as I have had several long back and forth dialogues with CoPilot about some of the ideas I have proposed in the book, testing them against different disciplines and imagined readers. But I am choosing to do the actual editing work with the help of humans, in dialogue and discussion with their expertise.

The obvious edits — then what?

Over the last two weeks, I have been trying to get through the obvious edits — the places I didn’t quite get the references right, the ‘insert quote here from that participant who…’, the ‘probably should have a footnote here’, bolded and highlighted sentences that indicate something is not quite working.

I have worked my way through those obvious edits, but I know that is not what is meant be editing. There is still some hard work to come. There are a few basic types of editing, and it’s worth knowing the difference. I might not need all of them at this stage, as the book will go out for review and no doubt that will tell me if there is others that need doing. I lay out the different types of editing below, as I understand them with reference to academic work.

Table 1: Types of Editing in Academic Work

Type of editing	What it means and when you might do it
Structural editing	Editing the structure of the book or sections of the book, including breaking up chapters, moving chapters, re-organising the steps in an argument. As a thesis supervisor, this one is usually required after the first full draft. It is also often required for books after review.
Developmental editing	Editing that focuses on the development of the ideas (or ‘characters’) in the book. Here we might look at pacing the introduction of certain ideas and arguments, how they develop over the book, and whether these developments make sense to a new reader. In my experience this often means reading through the book multiple times following certain threads or subthemes, and making sure they are not contradictory, and that have a logical flow.
Line editing	This is editing with an eye to flow, word choice, punctuation choice and so on. It might edit things that are not technically wrong, but that could be expressed better to fit the style of the book, the expectations of the audience or the publisher, and to sharpen any meanings that might be a bit too vague.
Copy editing	Copy editing focuses on style, spelling, grammar, punctuation and formatting. It is usually editing to a style guide. Here is where decisions might be made around what should be capitalised and why, whether the reference formatting is consistent with the guide, how numbers are expressed, which is the correct spelling for British or US English or whatever language the book will be published in. I usually hire a copy editor for a major piece of work — even though the publisher also uses one (I find contemporary academic publishers often outsource copy editing and I have had some terrible experiences of introduced errors).
Fact checking	Ideally, in an academic text, this should be done as you go — but it is still useful to think of it as a specific task. Read through the work, and ask yourself “is this true?”. It might be that things need looking up and correcting — what claims have been made, however small that might not accurate. It might be things like the year a conference was held that provoked a treaty, the disciplinary identity of an author you have mentioned (Kelly Dombroski, a feminist geographer working on issues of care in Asia Pacific, has argued…. But is Kelly a feminist geographer? Does she work on care in Asia Pacific? etc.) You might also check that quotes are reproduced accurately, that page numbers are accurate for the text referenced, and so on.
Proof reading	Proof reading is aimed at picking up minor typos, spelling mistakes, grammatical errors or word use errors.

After my sweep of obvious edits, my next step in my current book was to edit the introduction and the conclusion together. I like to do these together as it highlights if I have gone off track, written a different book from what I thought I was writing in the beginning. If they hang together pretty well, I can then edit the rest of the book to make sure the argument gets me from introduction to conclusion logically. So I would say I’m currently doing a developmental edit.

The first stage for me is to get a bit of space from the book, then read it through and mark it up as if I am supervising myself. I don’t do the changes, I just flag what the changes should be. This stops me getting drowned in the details of changes before I have figured out the overall scope of changes. This might require multiple reads, just as it does when supervising students — the first read through might be your initial reactions as a reader (wait — where did this concept come from? what is this about?) and your second sweep might be with a better sense of where the text is going and what might need to be done to make that work better.

Often for student work, or for shorter pieces of work, this stage is about cutting the number of concepts used to those that are necessary for the argument. For example, do I really want to introduce the complex history of a phrase like feminist political ecology if I really just want people to think about gender and power in the ecological relationships of the specific case study I am analysing? Depending on the audience, I might, or I might not.

I often find editing the book backwards is useful at this stage. Firstly because the earlier parts of the book have probably had more edits as I have often read through them while planning what to write next. But secondly, it is just a bit easier to see what setup is required for the conclusions made in each chapter by edit it backwards. You can do this two ways.

One way to edit backwards is to simply edit each chapter normally, beginning with the last chapter and working your way back to the first chapter. This is probably what I will do this time. However, another way is to edit backwards paragraph by paragraph. That is, you read the last paragraph and edit it normally, then move to the second-to-last paragraph, and so on. This really helps to see if the set-up for each step in the argument is present. I might do this later in the process — potentially at the copy-edit stage.

Making time for big-picture editing

My big struggle in all this is trying to find the mental and temporal space to do this big picture thinking and editing. The one-hour of pre-breakfast writing per day produced the first draft. But I find myself reading and re-reading during that same hour now and not being able to decided what to do. I follow up a few leads (“reference that book you have in the office”) but I haven’t really asked myself the difficult questions yet — or found a way to even figure out what those questions are.

So I am getting to the point where I am thinking a writing retreat is in order. It’s a terrible time of year to be thinking that, as I have the annual hui for our Rutherford fellowships, and I’m chairing a conference — and of course, all the kids’ end of year things will begin to crowd out the calendar. But the daily slog of student supervisions, meetings for research and for conference planning are crowding out my ability to think about my own project.

Writing retreats are a great way to clear the space of the calendar and focus on an important task for a sustained period of time. Even a three-day writing retreat can provide as much perspective and potentially actual work as a month of daily one-hour sessions on a project. Retreats — especially with others — are a great way to foster writing and research cultures that celebrate deep work. The idea is to combine focused work with a nourishing environment, one that helps reduce distraction and get the body into a calm, focused, attentive state. A view of the sea is ideal for me, somewhere quiet with places to recharge in a natural environment. But it can also be a well-lit public library, a boring internal room with no interruptions, or a noisy cafe with the passive supervision of others who may or may not be noticing what you are up to. Wherever we find ourselves attempting to retreat, the important thing is cancelling all the meetings we can, and creating space to think on our own piece of work.

Near the end of my last monograph, I was unable to go away anywhere for a writing retreat due to family commitments. Instead, I had a 4 week no-meeting at-home retreat! I had just started my new job at Massey, but had not physically moved there. I set up a temporary office in my laundry/mudroom space. I put together an automated message for a month, responding to all emails noting I was on a no-meeting writing retreat as I edited my book (as you may recall, it needed a lot of structural editing). I managed to get that difficult work done, although I ended up with a two week extension where I took meetings on Fridays only.

How will I do it this time? There is no way with my responsibilities that I could take another month without meetings (I recently went to Sweden and had no meetings). But I could potentially clear three to five days to go on retreat and try and iron out the biggest wrinkles I can see in the book. Watch this space — I will let you know!