There’s an enormous amount written (ironically) about whether to use AI in the production of text. Less is said about the world that text editors now inhabit.
I started my editing career as a very junior sub-editor on the Cape Times.
I was studying at the time. On Friday and Sunday nights I would arrive in the subs room at 6pm and work through till about 1am. This kind of piecemeal shift is called “dashing” in South Africa (supposedly because when you signed on for a shift, you put a dash in the roster instead of your name.)
I’ve done a variety of kinds of editing since then – and for years the basic assumptions were the same:
- A person or people write a text (in this post I will use the word text to represent the wide variety of things that might come across an editor’s desk: textbooks, blog posts, journalism articles, academic papers, novels, brochures… you get the picture.)
- Another person (or people) revise and change (that is, edit) that text.
- The text gets published (publishing can mean a range of things: printed as a flyer to go in people’s postboxes, or uploaded to a social media platform, or printed as a book and sold in a bookshop).
So far, so good.
In the last four years, since the advent of consumer-facing generative AI, many parts of that process have become fuzzy – largely because of the removal of humans from parts of the process. And those changes have profound implications for anyone who produces content now.
To get the full picture, let’s take a brief look at what editing meant up until November 2022, when ChatGPT was launched.
What editors did traditionally
I asked Google’s Gemini Notebook to summarise all the articles I’ve written on the subject (here). Here it is, edited and rewritten by me.
At a foundational level, editors refine text by resolving “mechanical” issues (spelling, grammar and the like) and improving structural coherence. Editing happens in levels, depending on the extent of intervention:
- Proofreading – rectifying spelling, grammar, punctuation, and inconsistent formatting, often according to a style guide (a list of rules about how text should be handled).
- Substantive copy editing – this might involve untangling wordy or clumsy phrasing, shortening sentences or constructing clear transitions (“bridges”) between disconnected thoughts. An editor might also look at chronological sequences, stepping in when narrative timelines become muddled or contradictory.
- Fact-checking – editors verify factual claims, check dates, test external hyperlinks, and confirm that statements make sense in context. (In journalistic and institutional publications, this includes identifying legal liabilities and issues of fairness.)
Finally, editors are collaborative partners in the process of creating a text. They are in a dialogue with the author, raising queries for clarity, and safeguarding the writer’s authority and voice.
The TL;DR: A human being makes a text. Human beings are fallible. Another human being helps iron out the fallibilities.
What we are now dealing with
In conversations with fellow editors, and looking the email list run by Professional Editor’s Guild , it’s apparent that texts “written” by generative AI are becoming ever more common. From student theses to non-governmental reports to actual books, much of the text we are asked to edit may have been created by a machine (even if that machine was initially extensively and cleverly prompted by a human being).
It’s universally recognised in the editing community that working with an AI-generated text is not the same as working with human-written text.
The work takes longer because it is not just editing; down-to-the bedrock rewriting may be needed. And the extent of fact-checking is greater. It’s common cause that AI hallucinates, and so nothing (nothing!) can be taken for granted. In the past, where a link to a source looked legitimate, and the sourcing habits of the author were known, it might not be necessary to click on every single link and examine the destination page. Now, every single link has to be investigated.
Above all, AI text is boring. And so the job of editing is boring.
What is an editor to do?
It’s probable that the people who produce AI-generated text don’t care much about the editor’s lot in life. But they presumably do care about the end-product. And to get a good product, the person doing the work needs to be fairly rewarded. That means that the apparent cost savings of using AI are offset by the need to pay more to an editor.
For editors wondering about their pricing structures, my colleague Gail Strauss proposes a triage approach. Her excellent suggestion, when faced with text that has obviously been AI-generated, is this:
First step: Establish if the writer or company care if readers think the text is generated by AI and may contain errors. If not, the editor just does the usual work, fixes some of the more obvious AI “tells” and invoices as normal.
Second step: If the client wants the text not to sound like AI, and wants it to be correct, charge them more, possibly double what they would have paid if they had written the text themselves.
What is a writer or content creation agency to do?
The choices are these:
- Use AI to write the text and publish it (cutting out the editor entirely).
- Use AI to write the text, and ask the editor to do what’s needed to tidy the text up and check for obvious errors.
- Use AI to write the text, and ask the editor to rewrite it along with doing a thorough fact-check, on the understanding that the work will take longer and will be more expensive.
Or – write it yourself (or hire a human writer) and ask an editor to do what they have been doing for generations. I guarantee you will get a better result in the end.
Main picture: A picture from 1915, labelled Book Manufacturing – Design and Editing. From Picryl / The Newberry Library. Licence: Public Domain
Previous Sensible Guide articles
Copy editing 101: Why a timeline matters – One of the most useful tools in a copy editor’s arsenal is the making of a timeline. Here’s how to do that.
Editing tips and tricks: Take nothing for granted – It’s easy, when you are copy editing, to get lulled into a sense that the words you are reading are as they should be. But the good editor’s motto is to take nothing for granted.
What does it mean when an editor raises a query? – Editing is not just someone correcting mistakes in something you’ve written. It’s a dialogue, in which the writer has a vital part to play…
How to be a good editor – it’s all in the routine – If I had serious money for all the articles and books I have edited over the years, I would be living in a rather nice beach house. Sadly, that is not the case – but I have amassed another kind of richness: a wealth of experience in The Way of the Editor. And the single most important thing about the Way of the Editor is establishing a routine
What you need to know about AI hallucinations (hint: they aren’t hallucinations) – When AI makes mistakes, it’s called hallucinating. There’s a better, if ruder, word for it.
If any of this is relevant to your work…
You might find one of these useful: