This article may contain affiliate links. Please see our affiliate disclaimer in the footer menu for more information. Thank you for your support!
If you’re considering becoming a proofreader, you may wonder if proofreaders are still needed. As a trained proofreader, I can assure you that proofreading continues to be a sought-after service.
People will keep using Grammarly, and now ChatGPT, for simple texts. However, proofreaders are still needed for complex, lengthy, or important documents. Proofreading software and AI cannot fully understand the complexities and nuances of language like humans; therefore, they can miss errors and sometimes make erroneous suggestions.
Let’s see why proofreaders are still needed and explore what kind of know-how technology would need for it to replace proofreaders.
Why Proofreaders Are Still Needed
Proofreading will be necessary as long as we continue to produce written content. Although proofreading software and AI are helpful tools, they’re no substitute for a skilled proofreader.
Only a trained professional can find the mistakes software and AI miss.
Will Technology Ever Replace Proofreaders?
People disagree about whether technology will eventually replace proofreaders entirely.
However, current proofreading software and AI can’t compete with a trained professional who not only understands grammar and mechanics but also the writer’s intent and voice.
Proofreaders understand the intricacies and subtleties of language in a way that machines do not.
They also work with style guides that contain specific (not generic) rules for correcting people’s prose.
Proofreading software and AI, while valuable, use algorithms that aren’t foolproof.
They still overlook blunders and make mistaken suggestions. We’ll look at specific examples of this in a bit.
Does ChatGPT Think It Can Replace a Proofreader?
Here is ChatGPT’s response when I asked if it thinks it can replace a proofreader.
Not entirely! While I can catch many grammar, spelling, and punctuation errors, as well as suggest improvements for clarity and flow, a human proofreader brings a keen eye for nuance, style, and context that AI can’t fully replicate. Proofreaders also understand tone, audience expectations, and subtle inconsistencies that AI might miss.
That said, I can be a helpful tool in the proofreading process—especially for first passes, quick checks, or brainstorming alternatives. Would you like me to review something for you?
Source: OpenAI, ChatGPT, 2025 (https://chat.openai.com)
Furthermore, in fine print at the bottom of the screen in ChatGPT, it reads:
“ChatGPT can make mistakes. Check important info.”
I also asked ChatGPT if proofreaders are still needed. Here’s its response:
“Yes, proofreaders are still very much needed! Even with AI-powered tools like Grammarly and spell checkers, human proofreaders play a crucial role in ensuring clarity, accuracy, and consistency in writing. AI can catch basic errors, but it often struggles with context, tone, and nuanced language issues.”
-ChatGPT 2025
Finally, I asked ChatGPT to explain why human proofreaders are still necessary. Here’s part of its response:
ChatGPT 2025
Here are all the reasons ChatGPT thinks we still need proofreaders:
- Understanding Context and Nuance
- Maintaining Voice and Style
- Handling Complex Grammar and Punctuation
- Catching Subtle Errors
- Understanding Industry-Specific Language
- Checking for Logical Flow and Coherence
- Fact-Checking and Common Sense Corrections
- Adapting to Audience and Purpose
- Handling Creative and Literary Texts
- Cultural Sensitivity and Ethical Considerations
ChatGPT agrees that we still need humans to proofread.
Now let’s delve deeper into why we still need these eagle-eyed error eradicators!
Software and AI Miss Too Many Mistakes
To illustrate how errors elude technology, I fed Taylor Mali’s humorous poem “The the Impotence of Proofreading” to ChatGPT.
I asked ChatGPT to proofread the piece according to the rules in one of the major style guides: The Chicago Manual of Style (CMOS). We’ll talk more about how proofreaders use style guides soon.
I provided ChatGPT with the entire poem for context. This wasn’t the first time I had ChatGPT proofread this poem by Mali; I had done the same in 2024.
Although ChatGPT’s performance has improved since last year, it still leaves a lot to be desired!
When I recently asked ChatGPT to proofread this poem, it said it recognized that the errors were intentional. Therefore, it only corrected the mistakes that it deemed necessary. Okay, ChatGPT, take it away. You’re the boss! 😉
Below is the response and ChatGPT’s attempt to make corrections.
We don’t have to look further than the title to see that ChatGPT has already made a mistake. Although it said it would keep the intentional spelling and grammatical errors, it took out the second the in the title.
The original title is “The the Impotence of Proofreading,” and the second the was definitely intentional.
ChatGPT also decided to put the title of the poem in italics. I’m not sure why. It’s the original poem, so the title should be left in roman font—the regular, non-italicized or bolded font style.
Furthermore, CMOS calls for putting poem titles in quotation marks, not italics, when referencing poems.
It also decided to put the obvious mistakes in bold font. Not that I asked it to do that; it just did.
Interestingly, it decided to keep the second the in the last line of the poem. Why take an extra the out of the title but leave one in the body of the poem? I’m not sure what goes on in the “mind” of ChatGPT.
Anyway, I wanted to see ChatGPT’s attempt at correcting the mistakes, even though they were made on purpose.
Here’s the beginning of ChatGPT’s attempt at fixing the fumbles. The confidence this chatbot has in its answers is interesting, isn’t it?
Let’s see the slipups that the AI model made.
ChatGPT Struggled with the Poem’s Title
Again, we see above that the title was incorrectly put in italics.
That’s a significant oversight from the get-go.
On to the next goof.
ChatGPT Made Countless Capitalization Errors
In the original poem, Mali only capitalized the first word of each sentence, with the exception of one instance that was likely intentional.
However, ChatGPT decided to capitalize the first word of every line.
Mali’s capitalization was likely a stylistic choice that a human proofreader would recognize.
A proofreader would query the author to ask whether their capitalization choices were intentional.
Only capitalizing the first word in a sentence is a poetic device called enjambment that creates enhanced rhythm and flow.
ChatGPT, you’ve taken a sledgehammer to Mali’s poem when only a chisel was needed.
It Allowed a Spelling Mistake to Slide By (Four Times)
The word spell chukker (from the original poem) should have been corrected to spellchecker or spell-checker, not spell checker.
CMOS follows the Merriam-Webster’s Collegiate Dictionary (11th Edition) for spelling. According to Merriam-Webster’s, spellchecker is correctly written as one word or hyphenated.
CMOS uses a specific dictionary to ensure consistency both within a single work and throughout the entire body of work.
ChatGPT left spellchecker as two words all four times it appeared in the text. Here’s one example:
It Made a Pair of Poor Punctuation Choices
In the original poem, no punctuation mark was placed after the word spelling (end of the second line in the screenshot above).
ChatGPT decided to insert an em dash after the word spelling.
Although some sources say that an em dash can replace a semicolon, CMOS doesn’t mention this in its section dedicated to em dashes.
Instead, it explains that these dashes are used to
“set off an amplifying or explanatory element and in that sense can function as an alternative to parentheses, commas, or a colon—especially when an abrupt break in thought is called for.”
-CMOS
I would have made those two independent clauses into two sentences or maybe used a semicolon, but not an em dash.
A similar fumble happened with the lines below.
In the original version, a comma was placed between the two independent clauses in the second line above. ChatGPT changed that to an em dash.
It seems like this chatbot is allergic to semicolons, which is what I would have chosen instead of the em dash.
A period could have worked too. But an em dash? Hmm.
It Likely Made an Error with the Teacher’s Name
Below, is part of the “corrected” poem ChatGPT spit out.
Did you ever have a teacher named Mrs. Myth? I sure didn’t.
As a proofreader, I’d ask the author if they meant Mrs. Myth or Mrs. Smith. And I’d almost certainly get a note back saying that Mrs. Smith was the correct name.
However, it’s important to note that proofreaders never make assumptions. Therefore, I wouldn’t change Mrs. Myth to Mrs. Smith without asking the author first. There’s a chance—albeit a tiny one—that the teacher’s name was, indeed, Mrs. Myth. You never know.
(ChatGPT doesn’t ask the author questions; it makes assumptions.)
Getting this name wrong is a fairly egregious error that likely would have been flagged by a human.
Now let’s delve into an essential theme: trust.
Not Everyone Trusts ChatGPT
I’ve been writing a children’s book for a while now, and I wouldn’t be willing to put it through ChatGPT.
Perhaps my concerns are unfounded, but I feel uneasy putting important information through AI due to concerns about privacy and information theft.
Any respectable proofreader will sign a nondisclosure agreement. However, ChatGPT isn’t able to do this.
Some People Want to Work with a Person
Would you rather collaborate with a person or a machine? I’d rather join forces with a person.
Some people will prefer working with people, even as technology improves.
Humans can provide insights and a connection that machines can’t replicate.
And humans care more about another person’s prose than software or a large language model does.
Proofreaders handle writing carefully, only making changes when necessary and never filling in the blanks based on assumptions about what the author meant.
As fellow humans, they understand the hard work and effort that go into creating an original piece of writing, so they’ll treat it respectfully and polish it to the best of their ability.
Companies Offering Proofreading Services Are Still in Business
Numerous companies still provide proofreading services, which is further evidence that proofreaders are still needed.
If you’d like to see some examples of these businesses, I wrote a comprehensive article about online proofreading work for beginners. Online proofreading and editing companies offer about one-third of the jobs mentioned in this post.
And if you’re unsure about the difference between proofreading and editing, feel free to check out this post on Om Proofreading.
The article discusses the differences between the four traditional editing stages: developmental editing, line editing, copyediting, and proofreading.
It’s important to note that proofreaders are sometimes the only professionals called upon to tidy up short texts.
Given the importance of proofreading, letting technology be the last set of “eyes” on a document is quite a crapshoot.
Proofreading Software and AI Don’t Understand Style Guides
One essential element that proofreading software and ChatGPT lack is the ability to account for all the differences between style guides.
We have countless style guides that provide standards for writing in various industries.
However, the four major style guides for proofreading English are the following:
- The Chicago Manual of Style
- The Associated Press Stylebook
- the MLA (Modern Language Association) Handbook
- the Publication Manual of the American Psychological Association
Except for the MLA Handbook, each guide specifies which dictionary to follow.
Proofreaders use different style manuals depending on the type of document they’re proofing.
I mentioned CMOS, one of the most esteemed, comprehensive, and commonly used style guides. It’s the go-to guide for proofreading books, and it can be applied to other types of texts too.
The rules from one style guide to the next can differ significantly.
Case in point, June Casagrande’s The Best Punctuation Book, Period, shows how each of the four major style guides differs in its application of punctuation.
It’s a handy resource for proofreaders!
Only a trained proofreader will know whether to accept or reject punctuation corrections made by technology since some modifications may apply to one style guide but not another.
And if you’re worried that style guides are a new trend that could quickly disappear, consider that CMOS was first published in 1906 and is now in its 18th edition.
An Example of a Snafu with Style Guides
Let’s say we’re proofreading together, and we come across this sentence:
Julie’s brothers—Ben and Tom—supported her in every step of her recovery.
Is this sentence correct? Well, it depends on where it’s written.
The punctuation in this sentence is acceptable if we’re proofreading a book. But it’s not if we’re proofreading a newspaper article.
Journalism typically abides by The Associated Press Stylebook, which almost always calls for space on either side of em dashes. Therefore, the sentence suitable for a news story would look like this:
Julie’s brothers — Ben and Tom — supported her in every step of her recovery.
Since ChatGPT (and Grammarly) aren’t programmed with style guides, how will they know when to leave the spaces on either side of the em dash and when to remove them?
How Grammarly Premium Dealt with This Punctuation Problem
I ran the above section of my blog post through Grammarly Premium (the robust version of Grammarly), and it failed to detect an error in either version of the sentence about Julie’s brothers.
It should have flagged at least one of the two sentences due to a lack of consistency.
These two sentences are fine as written in this article since I’m illustrating a point.
However, what if a client sent us a book manuscript that had em dashes with spaces on either side?
As proofreaders, we know those spaces have to go (per CMOS). But Grammarly didn’t let us know!
How ChatGPT Treated This Punctuation Pitfall
ChatGPT struggled also. Below is its response when I asked it to proofread the sentence—the one with spaces on each side of the em dashes—according to CMOS.
ChatGPT thought I used en dashes in the sentence I prompted it to proofread, but I didn’t. I used em dashes.
The chatbot produced the correct sentence but attached an erroneous explanation.
ChatGPT says it can provide guidance based on widely known CMOS principles. However, for niche or highly specific rules, it says the best practice is to consult CMOS.
(The AI model states that it doesn’t have direct access to the manual since it’s a copyrighted work.)
However, as we’ve seen, the guidance provided by ChatGPT can lead us astray.
Examples of Other Elements This Technology Doesn’t Consider
Checking that style guide specifications are met is just one example of what proofreading software and AI can’t do.
Let’s briefly look at a few other basics that are problematic.
An Inability to Check Cross-References
Cross-references refer to any part of a document that points you to more information located elsewhere within the same document.
Part of a proofreader’s job is to ensure these cross-references are correct.
For example, if a document reads, “See Figure 3,” does it refer to the correct figure? Or maybe the author meant for the reader to look at Figure 4, which contains the relevant information.
A Potential Blunder with Bibliographies
Now let’s think about the bibliography of a thesis.
What if the author lists a source in their bibliography but doesn’t cite it anywhere in their paper? During the editing stage, they may have scrapped the part of their thesis pertaining to that source.
A proofreader would know to eliminate that source from the bibliography, but Grammarly wouldn’t do that.
I doubt ChatGPT could do that, either. When I asked if it could, it didn’t give me a definite answer. It just reiterated that “this is something that should be flagged.”
Incompetence with Checking Where a Link Leads
As one more example, proofreaders of online content usually verify that links work and that they direct the reader to the right place.
This task goes beyond Grammarly and ChatGPT’s abilities.
Only a human can tell a website owner if they’ve linked to the wrong content.
Falling Short When Checking Formatting
Proofreaders watch out for formatting flubs.
Are indentations missing in some places? Is the spacing between lines and paragraphs consistent? Are page numbers in the proper order? Are there any font size or style changes where there shouldn’t be?
Let’s consider font style, specifically.
If you’re proofreading a book, the same font should be used to write “Chapter 8” as was used to write “Chapter 1.”
Proofreaders also need to consider formatting for elements (e.g., table of contents, index, appendix, bibliography, footnotes, endnotes) that may appear at the beginning or end of several types of documents.
In addition to checking cross-references for these items, proofreaders need to verify they’ve been formatted correctly and consistently.
Proofreading software and ChatGPT are limited in the kinds of formatting faux pas they can fix.
They would fail to fix formatting fumbles in our hypothetical book for obvious reasons—they can’t manage the sheer volume of text.
And if the book’s font style or size changed halfway through, it would look acceptable to ChatGPT. The language model converts all writing into a similar font when you copy and paste a document into the “Message ChatGPT” section.
Not Being Up to Date with Industry-Specific Terms
Although software and AI receive periodic updates, they’re not always aware of the latest terms that have become part of a particular industry’s vernacular.
They can also get tripped up on slang.
To wrap up, let’s see what proofreading software and AI would need to be able to do to substitute for a proofreader.
What Technology Would Have to Do to Replace Proofreaders
Grammarly or ChatGPT would have to be programmed with an immensely complex algorithm to be sufficient to stand in for a proofreader.
Here are examples of what it would need to know how to do:
- allow for breaking the rules occasionally
- correct mistakes in accordance with context
- accurately apply the appropriate rules in countless style guides and house styles
- comprehend humor and subtleties like wordplay and sarcasm
- consider writing within numerous genres and varying levels of formality
- have a firm grasp of a myriad of dialects
- query the writer when something needs clarification
- be able to check cross-references
- know how to fix a host of formatting flubs
- respect the writer’s style and voice, preserving the uniqueness of their work
- understand the writer’s intent and ensure the proper tone throughout a text
Furthermore, the algorithm would need to be tweaked whenever a dictionary or style guide was updated.
It’s hard to imagine proofreading software like Grammarly or a large language model like ChatGPT that can understand how to correct errors while considering context and style guides, all while preserving the writer’s style and voice.
A writer’s heart and personality continue to shine when their writing is in the hands of discerning individuals who prioritize respecting their authorial voice and creativity.
We don’t want to let an algorithm suck the humanity out of humans—especially when it comes to something as sacred as the delicate art of fine-tuning creative expression in the form of writing.
That concludes our conversation about why proofreaders are still needed.
If you’re interested in becoming a proofreader, you may want to check out my article “Become a Proofreader with No Experience (A 10-Step Guide).”
Best wishes to you!
“We don’t make mistakes, just happy little accidents.”
– Bob Ross
Recent Posts
Punctuation is important because it enables us to communicate our message clearly and effectively. Without punctuation, we wouldn’t understand how units of a sentence relate to one another or how...
Although you're probably somewhat familiar with adverbs, you may be unaware of sentence adverbs. As a trained proofreader who has studied the parts of speech, I can help you understand this unique...