Courtesy ChatGPT and Grammarly
Among ChatGPT’s myriad editorial skills is its ability to proofread your text at an extremely high level of proficiency.
And if all you’re looking for is an occasional proofread, you’ll find you can get help from ChatGPT — combined with similar AI tools — for free.
I began using ChatGPT as a back-up proofer for my primary proofreading tool — Grammarly — after I saw it consistently catching errors in text that Grammarly was missing.
Granted, the errors ChatGPT catches are generally minuscule (Grammarly really is an excellent proofreading tool, among its many other charms).
But sometimes, ChatGPT does catch outright misspellings that simply fly by Grammarly.
And on average, I find that ChatGPT spotlights about one-to-three errors that are missed by Grammarly per one thousand words of text.
One of the great beauties of creating a proofreading prompt for ChatGPT (a sample of one of mine follows) is that you can design it to do proofreading exactly as you prefer.
My proofer prompt, for example, instructs ChatGPT to retype the entire text it’s proofing, highlight each error it finds in bold, offer a suggested correction — and then explain its rationale for making each correction.
That may be overkill for some people.
But I prefer double-checking ChatGPT’s work myself and being given a sound reason for making a change — rather than simply trusting ChatGPT to ‘correct’ my text with no oversight.
I also greatly appreciate that I can add new rules to my ChatGPT proofer — or take some rules away — by simply changing a few words in the prompt.
And I really like the fact that my proofer grows more powerful with every upgrade of ChatGPT’s software.
For example: The ChatGPT proofer I use most (below) is more powerful and more meticulous in its proofreading when used with the most advanced version of ChatGPT — ChatGPT 4o — as compared to its use with previous versions of ChatGPT.
As with many things ChatGPT, your success designing a proofer hinges mightily on the precise wording of your prompt.
Sure, you can design a proofer with something simple like: “You are an expert proofreader. Please proofread the following text and highlight the errors in bold.”
But I’ve had better luck using a longer prompt that spells out — in excruciating detail — exactly what I’m looking for.
For example: In addition to asking for an error scan, I also ask ChatGPT to be meticulous as it’s scanning, to focus intensely on its task, to get extremely granular in its inspection (such as identifying correct placement of periods) and similar.
But simultaneously, I’m also careful not to overload ChatGPT with too many different kinds of requests in a single prompt — which can needlessly slow down ChatGPT, or even confuse it entirely.
As previously indicated, if you’re only looking to do some occasional proofreading, you can do so for free using ChatGPT and other tools and still enjoy a high level of performance.
For example: You can sign-up for Grammarly’s free version, which will do a basic, highly effective proofread of your text at absolutely no charge.
And then you can head over to ChatGPT, where you can double-check Grammarly’s work using free, limited access to ChatGPT’s most advanced AI engine, ChatGPT 4o.
Granted, you can only input about 15 prompts every three hours using ChatGPT-4o’s free version.
But if you only have a few documents to double-check for proofing, you’re all set.
For more extended, free proofing, you can also use Grammarly in combination with Gemini Advanced — Google’s direct competitor to ChatGPT-4o.
The reason: Gemini Advanced is based on the same genre of AI technology that powers ChatGPT. So a prompt developed for use with ChatGPT will also work flawlessly with Gemini Advanced.
Plus, Gemini Advanced currently offers an extremely generous, two-month free trial.
So chances are, you can theoretically proof an entire book if you so desired — over time — just using Grammarly’s free version, in concert with the Google Gemini Advanced’s free trial.
Another bonus: Given that the lion’s share of AI writers/tools on the market right now use the same genre of AI technology as ChatGPT and Gemini Advanced, you can also use any prompt initially developed for ChatGPT in any other AI tool based on the same genre of AI tech, including:
A final tip before we take a look at a sample proofer: It’s always best to design your proofer as a stand-alone prompt — followed by a second prompt that you use to actually input the text you want proofed.
This technique ensures that you’re not giving ChatGPT too many words and/or too much data to process in any one prompt.
Even better: Inputting your text in a second prompt also expands the number of words ChatGPT will proofread for you, given that ChatGPT can only process and understand so many words in any given prompt.
In any event, below is one of the homespun proofers I use when I’m proofing with Grammarly and I’m looking for ChatGPT to do an extremely granular double-check of Grammarly’s work.
One caveat: Due to ChatGPT’s inherent nature to often respond to the same prompt in a different way, it may ignore one or more requests of any prompt it’s given. If this happens, a slight rephrase of the prompt usually corrects ChatGPT’s oversight. Here’s the prompt:
Joe Dysart’s Proofer Prompt:
[You are a seasoned, award-winning expert in proofreading. Please conduct a meticulous, line-by-line proofread of text that I will provide you in the next prompt.
Focus intensely on the following aspects:
*Technical Accuracy: Identify and correct errors in spelling, grammar, punctuation, capitalization and formatting.
*Other Technical Errors: Identify and correct other technical errors not specifically listed in this prompt.
*Correct Placement of Periods and Quotes: Pay special attention to the placement of periods and quotes. Identify and correct any incorrect placement of periods and quotes and incorrect spacing associated with them.
*Consistency: Identify and correct inconsistent use of terminology, names, spelling, formatting and internal style rules (if applicable).
*Specific Conventions: Verify that the following conventions are employed:
~Allow the use of slang
~Allow the use of sentence fragments
~Allow the use of sentences that start with And
~Allow the use of sentences that start with But
To report the errors you find:
*Preserve Original: Retype the entire text with your changes.
*Highlight Errors: Highlight each error you find in bold.
*Provide Correction: Offer your correction for each error in parentheses directly after the bolded error.
*Explain Correction: For each correction you make using boldface type, please explain the correction, also in boldface type, using the format in the following example, which is delineated by three quote symbols “””solider (soldier) — The word soldier was misspelled.”””
*Always keep in mind your overarching goal: Your overarching goal is to provide the most comprehensive, meticulous proofreading of the text as possible.
Please indicate that you fully understand these proofreading instructions and have stored the instructions with the statement: I fully understand these proofreading instructions and have stored the instructions. Please input the text you’d like proofread with your next prompt using the words, “Here’s the text to proofread:”]
Share a Link: Please consider sharing a link to https://RobotWritersAI.com from your blog, social media post, publication or emails. More links leading to RobotWritersAI.com helps everyone interested in AI-generated writing.
–Joe Dysart is editor of RobotWritersAI.com and a tech journalist with 20+ years experience. His work has appeared in 150+ publications, including The New York Times and the Financial Times of London.