by Loretta Bushell, a writer for Reedsy
Artificial intelligence is the new pineapple on pizza: some people love it and others hate it, but everybody has a strong opinion about it.
On Reddit, I’ve seen people who think human proofreaders will soon be redundant, and I’ve seen people who insist that going anywhere near AI automatically makes you a “lesser” author. The reality is somewhere in the middle: AI can be a useful self-editing tool, but it can’t replace a professional proofreader. Read on to find out why.
AI can help you self-edit
Before any of us had ever heard of AI, we all used spell checkers to catch typos and missing punctuation marks. Crucially, this was merely one stage of the proofreading process — something to do before hiring a proofreader, but certainly not something that could perform all the same tasks.
AI serves much the same purpose. Sure, it can also make stylistic suggestions to help you reword clunky phrases and improve your sentence flow. But it’s still automated and, to some extent, limited by the prompts you give it.
The biggest advantage generative AI holds over normal spell checkers is that it’s more customizable, albeit not always in a cumulative way; that is, it doesn’t necessarily “remember” how you instructed it previously. However, you can use an individual prompt to tell AI precisely what to check for, defining your own personal style guide. The more detailed your prompt, the better a proofreading job it will do.
Here’s an example of a prompt you might use when self-editing sections of your manuscript:
You are a proofreading expert whose job is to correct grammatical errors in the text provided. Your task is to analyze the text, identify any mistakes, and make the necessary corrections in bold. Please check for spelling, grammar, and punctuation errors, as well as correct word usage, sentence structure, clarity, and flow. Please use US spelling and punctuation rules and apply the Oxford comma throughout. Don’t correct the passive voice or excessive adverb use.
Please add a ‘Summary of corrections’ section after the text that lists all the changes made, along with a brief justification for each.
Using AI in this way can enhance and speed up the self-editing process. But just as you wouldn’t make do with a spell checker alone, you shouldn’t only use AI to proofread your work — because both are known to skip over critical errors.
AI can make mistakes
To be fair to AI, humans can also make mistakes — but decent proofreaders would never make oversights quite as embarrassing as the ones artificial intelligence does.
As well as potentially missing errors that would be immediately obvious to a human reader, AI can actually introduce mistakes into your manuscript. For example, if it comes across an unfamiliar technical, regional, or foreign term or expression, it might “correct” it into a word or phrase that it is “familiar” with.
Here are just some of the proofreading mistakes AI has been known to make:
- Removing words “for concision” in a way that alters the meaning of the sentence.
- Miscorrecting words that are already correct. For example, writer G.B. Pool reports that her AI spell checker turned “everything was socked in” into “everything was shocked in.”
- Failing to recognize a misspelt homophone (e.g., “bear”/”bare”).
- Not spotting inconsistent spelling of the same proper noun.
- Failing to capitalize uncommon proper nouns.
- Leaving awkward phrasing in — or again, even introducing it!
The bottom line is that you can’t trust AI not to make mistakes. While it can help you identify some errors, you’re still going to need a human to proofread its output.
It flattens your voice
The potential inaccuracy is not even the worst problem with letting AI proofread your work. While it will likely get better at identifying more “nuanced” mistakes over time, AI is always going to lack the ability to not correct deliberate “mistakes” and originality.
By its very nature, AI identifies and follows established patterns. Given the chance, it will flatten your voice until you sound more like everyone else and less like you, by “correcting” your creative language and forcing it to adhere to certain grammatical rules. No matter how you instruct it, AI will almost always fall back to doing this — which is a shame, because intentional “mistakes” can have an incredibly strong stylistic impact.
To illustrate this point, here are some authors whose writing would be labeled as poor or incorrect by AI:
- James Joyce uses a sentence that is 3,687 words long in Ulysses to create a unique stream-of-consciousness effect. For context, the AI spell checker LanguageTool marks any sentence over 40 words long as a mistake!
- In chapter 1 of The Brief Wondrous Life of Oscar Wao, Junot Díaz writes, “he wasn’t no home-run hitter.” An AI proofreader would replace “no” with “a,” losing the slang that characterizes the narrator’s voice.
- Toni Morrison deliberately spells “blackman” and “whitelady” as one word each in Beloved, to show how people are viewed according to their racial identity. AI lacks the contextual understanding to see that this “mistake” is purposeful.
- In The Snow Child, Eowyn Ivey omits quotation marks in dialogue between the Snow Child and other characters to emphasize her otherworldly nature and make readers question whether she truly exists. Again, AI would be unable to make this link.
AI also runs into problems when it comes to creative puns. I asked ChatGPT to proofread the introduction to an article I’d written recently:
You don’t have to plot your novel before you start writing — just like you don’t have to get breakdown cover for your vehicle. But you’ll regret your decision if a massive p(l)ot hole derails your journey halfway in!
ChatGPT told me to remove the “(l)” — indicating that it didn’t understand the punny intention of my paragraph at all!
Nothing beats the human touch
As shown, AI doesn’t really understand you… but a good human proofreader will. Thanks to their, well, human instincts, professional proofreaders can easily figure out when an unusual choice has been made for a specific reason — and if they can’t, they’re able to listen to your explanation with emotional intelligence and determine the best way forward.
But it’s not only your decision-making process that AI can’t comprehend; it also doesn’t understand what it’s like to be a writer. A human proofreader can acknowledge the huge amount of time and effort that has gone into your manuscript, offer meaningful encouragement, and deliver any constructive criticism sensitively and at the right time (i.e., not when you’re in the middle of a crisis and doubting your whole career).
Professional proofreaders also care about protecting your confidentiality and will never recycle your words or style on another client’s project — which obviously cannot be guaranteed if you run your chapters through AI.
Ultimately, there are four questions to ask yourself when choosing between AI and a professional proofreader:
- Do I want to make sure my manuscript is free from embarrassing mistakes?
- Do I want my unique creativity to come through in my writing?
- Do I appreciate the value of human connection and empathy?
- Do I want to protect my work from being copied without my consent?
If you answer “yes” to one or more of these questions, then you need a professional proofreader — and that’s not going to change any time soon.
***
Loretta Bushell writes for Reedsy, a marketplace that connects authors with resources and professionals to help them publish their dream book. Loretta is also a part-time translator and spends her spare time reading books of all genres (except horror!) and playing board games.