The methods I use to edit my writing with the help of AI.
Did you know it’s very difficult to proofread your own writing? That’s because you already know what you meant to say, so your brain fills in the gaps automatically. As a result, it tends to overlook the actual words and letters written on the page.
Every writer has experienced the frustration of proofreading a piece multiple times, only for someone else to immediately spot a glaring mistake.
I once paid $80 for a box of business cards. I must have proofread the card a dozen times before sending them to print. When they arrived, I eagerly showed them to my wife. But as soon as I handed her the first card, she glanced at it and gasped; I had spelled the name of my business wrong. No matter how many times I checked, my mind kept seeing what it expected to see, completely missing that two letters in the first word had been reversed. In the end, that entire $80 box of business cards went straight into the garbage can.
I use Microsoft Word for my writing, and its spell-check and auto-correct features are invaluable. However, there are several critical editing tasks that Word cannot perform (at least not with the default installation I use). Its spell-check and auto-correct functions will not identify any of the following writing mistakes:
Run-on sentences, sentence fragments, comma splices, misplaced or dangling modifiers, subject–verb disagreements, shifts in tense, shifts in point of view, homophone confusion (e.g., right, rite, write, wright), incorrect prepositions, redundant pairs, wrong plural or possessive forms, using adjectives instead of adverbs, missing commas, improper paragraph breaks, and more.
But a language modeler like ChatGPT can catch all of these mistakes, and more. Every good writer needs someone to proofread their work. However, giving your writing to a random friend or family member offers no guarantee that they have the skills or training to spot these errors. Hiring a professional editor, meanwhile, can cost thousands of dollars and take weeks or even months.
An AI program, on the other hand, can identify every type of writing error; and do so in mere seconds.
But here’s the catch:
AI programs like ChatGPT will always take bright, impactful, and colorful writing and make it bland, boring and dull. For this reason, every writer who uses AI as an editing tool must also learn how to “edit the editor.”
The following is my personal list of how I use AI as an editing tool for my writing.
#1) Every sentence and every paragraph must be my own original thoughts.
When I was in high school, using a calculator on a math test was considered cheating. Any student caught doing so would receive a zero on the test and a call home to their parents. College, however, was very different. In my engineering classes, using a calculator during tests was required; but each student had to have memorized the formulas. Typing numbers into a calculator is meaningless without knowing which formulas are needed to solve the problem.
In high school, using a calculator was cheating because it allowed the student to sidestep the work and let the calculator do it for them. In college, however, the calculator became a tool to display the results of the student’s knowledge.
Likewise, if a person uses an AI language modeler like ChatGPT to bypass the work of putting their own thoughts on paper, they are cheating. They are being dishonest. But if a writer uses ChatGPT as a tool to make the words on the page more accurately reflect their own thoughts, that is professional and mirrors exactly what a human editor would do.
Actually, let me revise that last point. I once wrote for a magazine published by my church’s denomination, which was distributed nationally to all the churches in that denomination. I had a human editor, and I vividly remember the first article I submitted. My editor made so many changes and rewrote it so extensively that upon reading it in the published magazine, I felt guilty being listed as the author. I didn’t recognize a single paragraph. I felt like a fraud and thought it would have been more appropriate if the editor’s name had appeared as the author.
By using an AI editing program, a writer can ensure that the thoughts which originated from their own mind are preserved on the page. Writing errors can be corrected, and the arrangement of words can be refined so the original idea is expressed more clearly and precisely. In my opinion, using an AI editor can be even more effective than a human editor, because I, the writer, know exactly what I intended to say. I remember what I was thinking when I put the words on my computer screen, and I can guide the AI editing process to convey my thoughts accurately. A human editor, however skilled, cannot read minds and therefore cannot fully know what the writer’s original thoughts were.
I feel very fortunate that, in ninth grade, I was required to take a typing class. At the time, I didn’t like it, but as an adult with an interest in writing, being able to type with all ten fingers has been invaluable to me. I genuinely feel sorry for those I see typing using only their index fingers and thumbs, a method they likely learned from operating a video game controller. Can you imagine writing an entire book that way? At 10 or 12 words per minute, it would take years. For me, the best way to start a writing project is to sit down at my laptop and free-flow my thoughts onto the screen, typing as fast as I can think. Those thoughts then form the foundation for what is later edited.
#2) Emotion is what makes good writing
The word “poignant” means to evoke a keenly felt sense of emotion. If your writing contains any poignancy; which is what makes good writing; you can be certain that ChatGPT will try to generalize it to the point that your writing loses all of its emotion. While it may correct a run-on sentence, it can also make your writing dull in the process. As the writer, and the “editor of the editor”, you must restore your poignant expression while also recognizing the valid corrections it made.
I always place the edited output side by side with my original and compare them word for word to see exactly what was changed and why. I might make several changes to the output and then run it through the editor again, repeating the process until I have precisely conveyed my original thought.
#2) Never ask ChatGPT to edit more than one paragraph at a time.
If you write an article and then submit the entire text to ChatGPT for editing in a single copy-and-paste operation, it will likely end up destroying your piece.
Imagine hiring a handyman and telling him, “Here’s my house; fix what’s wrong.” You leave to go to work, and at the end of the day, you come home to find that he has converted your entire house into a duplex. Now your home has two front doors, and strangers are moving in.
Instead, you should have brought the handyman into your home and said, “The sliding door on this closet doesn’t shut properly. Please fix only this door.” A few minutes later, you would find that he fixed it, and that it works perfectly.
By giving ChatGPT only one paragraph at a time to edit, you greatly limit how much it has to work with. As a result, the extent of its alterations is also limited. It can correct your comma splices and adjective use, but it can’t change the overall direction and meaning of your writing. You simply haven’t provided it with enough content to do so.
#3) Don’t let ChatGPT make changes just for the sake of changing something.
ChatGPT will often change words simply for the sake of change, not because the revision improves the writing. That’s why you may often desire to put your words back into the text. The thoughts that originated in your mind are what you want your audience to read. When ChatGPT alters words unnecessarily, it strips the “you” from your writing. The goal of using ChatGPT as an editing tool is to ensure that your own thoughts are clearly and accurately conveyed to the reader. Correcting genuine writing mistakes is important, but making changes just for the sake of change is not.
#4) Remove the AI signatures.
I’m not sure why, but certain words, punctuation choices, sentence structures, and writing styles are so heavily favored by ChatGPT that they have become hallmarks of AI-generated text. The word gentle is one example: “Would you like me to gently recommend a pizza recipe?” “The gentle nature of her suggestion.” “God’s gentle love for His people.”
This really ties back to my previous rule. ChatGPT will edit your writing by inserting words, phrases, and punctuation that you, as the writer, would never use. All of that must be stripped out and restored to reflect the way you actually write.
For example, most people don’t know how to type an em dash (Alt + 0151). However, ChatGPT uses em dashes as liberally as I use salt on scrambled eggs, making them a signature of AI writing. When “editing the editor”, I always replace em dashes with my preferred punctuation; typically, a comma or a semicolon. Much to my consternation, however, I’ve noticed that the Berean Standard Bible, which is what I always use when quoting Scripture and which was published long before the first AI programs existed, frequently employs em dashes. I have to remain constantly vigilant to avoid accidentally “editing” Scripture while reviewing my own commentary writing after using an AI editor.
The obvious goal of writing is to convey one’s thoughts onto paper. Poor writing, however, can cause the reader to walk away with unintended messages. A humorous and often cited example is the phrase “Let’s eat Grandma.” Clearly, the author did not intend for his grandmother to be eaten. The correct sentence should have been “Let’s eat, Grandma,” or better still, “Grandma, let’s eat.”
Every writer needs an editor because it’s extremely difficult to proofread one’s own work. I once bought a book, brought it home, and began to read it. But somewhere in chapter five, I noticed that the author had used the word “containment” where the context clearly required the word “consignment.” I went to the author’s website and emailed him about the mistake. He replied with gratitude and admitted that four human editors had failed to catch it.
The incredible value of using an AI editor is that it’s virtually impossible for the machine to overlook such subtle errors. The machine doesn’t misspell words or make writing mistakes. ChatGPT was trained on about 570GB of text, roughly 300 billion words, or the equivalent of six million books. It is, for all practical purposes, impervious to grammatical errors. While it obviously lacks the human component that we, as writers, must restore to its editing outputs, ChatGPT is, in my opinion, the most effective editing tool available; provided you’re not lazy and are willing to first record your own thoughts and then afterward, take the time to edit the editor.