You may have heard that artificial intelligence is going to revolutionize writing and other creative endeavors. That it is now possible to use AI to both create and edit text. As editors, we believe that, while AI can be useful in a few contexts, there are good reasons to be on the lookout for its use in places where it is unhelpful, or even harmful, to your writing goals.
In this post we outline the basics of AI as it intersects with editing. We’ll tell you what we believe about its place in the landscape of freelance editing. By the end you will be able to make more-informed decisions about the application of AI to your writing—be it a book, article, thesis, or short story.
AI Primer [skip]
What is AI?
So-called artificial intelligence (AI) is software that mimics a human or animal brain activity, and is programmed by “training.” It returns an output when given an input—a prompt (for example, “write a cover letter for a book proposal”). Software that uses AI algorithms does not have general competence: it can do only things it is trained to do.
AI software for writing and editing uses a Large Language Model (LLM). The text LLMs output is determined by a sort-of-average next word that might follow the previous word, or likely sentence that would follow the previous sentence. LLMs work probabilistically, so given the same input they can return unique outputs. It says the kind of thing someone would say.
That “average,” so to speak, comes from some large corpus of writing (for example, every post Facebook has saved). Such an AI necessarily produces a mediocre product (for example, if it writes like an average-ish user of a large social media platform). The most sophisticated LLMs use more or less every piece of text that has been digitized.
Recent releases of AI programs are measured against benchmarks in areas such as writing computer code and reporting scientific facts: strictly rule-governed activities with clear validation procedures and large bodies of high-quality examples to “train” on.
Where is AI?
AI is being integrated into operating systems and apps widely, if not uniformly. Those using Windows, Android, and Office products will have noticed updates introducing them. It is also used as a tool for image and video creation, personal planners and simple scripted web-based tasks, voice recognition, and so forth.
AI can be used as an aid to composition. Some apps, like Grammarly (to use a widely known example), advertise this function. LLM-based AIs can swiftly generate large quantities of text.
It is also used for editing. Here are three prominent uses:
- A “smart” grammar and syntax correction “assistant” built-on to a word processor. This is a version of familiar tools long available in the Office suite of programs, for example.
- A “smart” grammar and syntax correction standalone app. This might be a familiar app that interacts with a word processor like Word, such as the aforementioned Grammarly.
- Inputting a text directly into a ChatGPT-type interface and directing the AI to correct or evaluate it.
There are other writing-adjacent areas AI may crop up, for example reading blackletter manuscripts from scanned images or serving as an aid to translation in a lit review context, that are not relevant to our purpose here.
AI in the Freelance Editing Business
Our perspective
In short, we do not approve of, and do not use AI in professional editing contexts, except for very low-level error-detection. Expert human knowledge of how to apply and allow exceptions from customary rules (e.g. from the Chicago Manual of Style), and the ability to detect subtle and contextual semantic details in a specific text can make a strong impact on a reader’s reception of a text. We believe in the value of our clients’ work and see our responsibility as increasing that value.
App assistants and duplicated work
Assistants to editing are fine, in our opinion. We use some automated tools. Where one trips up is trusting too much in a product that is only “good enough” for it to be sold to the public. You might suspect that the insertion of AI into many, many apps, and it being aggressively marketed as the next big thing is, in fact, hype. We would tend to agree. In any case the fact is you must trust or be able to confirm that an app’s corrections are effective and do not introduce defects. Human eyes need to check the grammar-checker.
When self-editing, this is not such a big deal — for example, updating a highlighted “problem” sentence, or choosing from a list of suggested fixes to one, is not nearly the same level of investment as needing to approve already-made alterations to a text. In contrast, when you need an outside eye on a text that you’ve been working with for weeks or months, even a tiny rate of introduced error matters a great deal. Using AI for whole-text revisions tends to create duplicative work.
Unwanted changes
When an AI alters a text with freedom to re-write, it may introduce unwanted changes that are beyond the scope of the services requested. AI has been known to fabricate citations, for example. It might treat direct quotations like the surrounding text in some cases. An editor using AI may or may not check for and undo such changes. If the job requested is a “copy edit,” for example, this produces problems that require additional time to detect and fix.
The quality of AI-assisted editing varies
Some freelance editors use AI assistants to replace some of their work. Literally accepting all of Grammarly’s suggestions reflexively, or even telling a program like ChatGPT to “fix” your text. This does not always do justice to your voice and the intended meaning of a text, and in some cases creates serious mistakes. There are enough stories of texts that were outright mutilated by what are essentially chat bots—coming from authors—that we must mention it.
Others may use an AI assistant only to identify potential problems. When working with a checklist for copy-editing, for example, this might increase efficiency without changing how the editor produces the marked-up/corrected text.
In another type of case, an editor might ask an AI to evaluate a manuscript, essentially writing the bulk of an editorial letter. As you might expect, the results will tend to be bland, generic, or vague, be driven by large-publisher marketing concerns, and ignore self-contradiction.
Market pressure and AI
The freelance editing space is saturated across the range of price points. AI allows a desperate or unscrupulous person to “work” at a high rate of speed while competing on price. The need for short turnaround, given low hourly rates, means that the returned product attains to a lower standard of quality. In short, these editors tend not to do enough checking of the outputs of their software.
It is always worth considering how you came across information about a freelancer, and whether they will answer questions about their process in this regard, your risk-tolerance, and whether you want a one-time service (compare: raking and removing yard waste) or an ongoing relationship (compare: your trusty horticulturist).
Further considerations: What do you want editing to do for you?
Different writing projects have different needs. For example, you might want check-ins and face-to-face coaching; you might have to make a conference talk “good enough” on a short timeline; you might need a whole manuscript evaluated in a specialist subject area and with knowledge of theoretical texts; you might need help finding your artistic direction with a creative manuscript that you’re “stuck” on. Because of its lack of higher-order judgement, its lack of ability to apply knowledge to new areas, its inflexibility at the boundaries of what it is trained to do, AI tends to be a poor fit for subtle, specialized, innovative, personal, creative, and humanities-research projects.
AI editing/writing tools can be good for routine and mundane tasks that simply need a basic level of communication, such as writing administrative emails. Or for following validatable rules, such as creating small code blocks. But as you look around the internet, you’ll see vast swathes of bland slop articles that are produced at high speed by AI. If your work is precious to you, taking your text in that direction is not what you want. For ourselves, we always want your text to feel right.
In addition, perhaps you need a person who can work with you to make choices about what kind of support is best for your manuscript. (You are most likely not a professional editor yourself.) Knowledgeable advice about in what order to make revisions or purchase editing services that work with your situation, budget, and deadlines can make your options much more clear. Being able to talk to a person who will tailor their work to you, the author, is invaluable.
Stakes, stakeholders, and responsibility to clients
Your writing is personal to you. We believe that means it is valuable both for its content and because of its relationship to the author. Whether the text achieves your writing goal is a minimum. A manuscript should represent your voice, knowledge, choices, intention. The stakes are high. We therefore want to handle it with care.
We also believe that we should be at least somewhat satisfied by the results of our efforts. We are, in a small way, stakeholders in the final text. So, we want to focus on writing projects and practices to which our skills are applied with the best effect. From developmental and coaching-writing contexts to copy and line editing, we believe that it is our responsibility to elevate the quality of a text while honoring the needs and desires of the author, and without leaving (detectable) fingerprints on it.
Your editing needs may differ
Sending your work to an outside editor is an exercise in trust. That’s why we use consultations to get to know one another a little before beginning work, as well as to better understand how we can tailor our services to you.
Finally, here’s the part where we urge you to get in touch with us about your own writing projects. [Link]