When you’re faced with writing deadlines and you know you have to proofread your work before sending it out for review, it can be tempting to rely on ChatGPT or another artificial intelligence (AI) tool as a free (or inexpensive) in-home editor.

I’m here to tell you: THIS IS A TERRIBLE IDEA!

Yes, I know, I’m an academic editor so I have a huge stake in this debate. BUT that’s not why I’m opposed to using it (of course, there are also ethical and copyright issues with using these tools, but those are beyond the scope of this article). Rather, using AI as an editing tool is an awful plan because it does a horrendous job of editing text. I’ll give you three reasons here, including specific text examples; make sure you read to the end because the last reason is not what you’d expect!

Note: For this article, I used ChatGPT 3.5. I know there are more sophisticated options out there, but I assume that most people are using free tools and that ChatGPT 3.5 (being the most talked about) is the one you’re most familiar with. Let me know in the comments if I’ve gotten something wrong or if there are other (free) tools that can do the things that GPT 3.5 can’t.

1. ChatGPT’s “knowledge” is limited.

If you’ve been following news and information on AI, you know that ChatGPT’s “knowledge” is limited to the text that it’s been fed during its training. For example, I asked Chat GPT “When does Chat GPT’s knowledge stop?” This was its reply:

As of my last training cut-off in January 2022, I don’t have real-time updates or the ability to access information beyond that date. I can provide information and assist with a wide range of queries up to my last training point, but for the most recent developments or events, you should verify with up-to-date sources.

While it has a lot of information to pull from, January 2022 is now two years ago (but see this article that creates even more confusion). So much scholarship—for example, in politics and international relations, health, business and finance—is based on information that is constantly developing. Anything that you’re writing about that relates to events of the last two years is not going to be assessed by ChatGPT. This includes, just to name a few:

• the Israel–Hamas war, the war in Ukraine;
• COVID: updates on disease variants and immunizations;
• international policy related to the UN, WHO, IMF, and so on;
• election results; and
• business and finance: new startups, the constant chaos that is Elon Musk, cryptocurrency updates.

Meanwhile, a human editor likely follows the news, knows how to do an internet search to fact-check information, and even keeps up on the latest research simply by virtue of the fact that we read so many articles and books that are being submitted for publication. This last point is especially true of editors who specialize in particular fields or subfields.

2. ChatGPT is often WRONG.

First, ChatGPT and other AI programs lack subject-specific knowledge that a human editor (with a PhD – Hi, that’s me!) has. When it reads your text, it also can’t stop and think, “Hmm, I don’t think that’s quite right. Let me look that up and double check.” Even though you (the author) are an expert in your field, it’s easy to mix up dates or terms—particularly names of organizations, policies, or even authors you’re citing—and to overlook these small errors during your own writing and proofreading. But a skilled and experienced editor, especially one who specializes in your field, is much more likely to catch these mistakes. Many authors make the same errors, so we editors see them all the time. We know when something sounds “off” and needs a quick Google search to double-check.

As an example of ChatGPT’s imperfect editing, I copied the first paragraph of my dissertation, added some grammatical mistakes and typos, and asked ChatGPT to edit the text. This is what happened.

Text I entered vs. what ChatGPT gave me:

Explanation:

While the program caught several of the basic grammatical errors (missing commas and a hyphen, in red), I would argue that the wording changes in yellow are not better than the original text (IMHO), and the red-highlighted phrase changes the meaning of the last sentence and is incorrect (Kennedy wasn’t a moderate or non-ideological, he was very liberal!). ChatGPT also did not change the red hyphen in the year range to an en dash, which is required by almost every publisher and citation style.

Here’s the rest of the text.

Explanation:

So this one is a bit more complicated. We can quibble about whether or not the yellow rephrasing is better or worse than the original, but ChatGPT made some clear mistakes:

1) It missed the need to add a second quotation mark in the parenthetical citation (though it did add the comma in green), and I would hyphenate “roll-call voting”;

2) The blue text is fine, but a simpler revision would have been to add the “ed” to make it “crossed party lines” (this was an error I introduced to the text for the sake of this example); and

3) The program doesn’t seem to be very good at avoiding passive voice or redundancy: “resulting in the passage of” is much less direct than “to pass,” and “throughout his entire career” is redundant—“throughout his career” would suffice (this was another error that I introduced).

Now, you might be thinking, “This is tedious. Who cares?” But if you’ve ever submitted a paper to a journal before, then you know that some reviewers take issue with these problems and will reject you simply because your article is poorly written. Errors and bad writing also make it harder for the reader to understand the core of your argument and the evidence you present; when they expend mental energy trying to understand what you mean, they lose interest and can’t focus on the significance of your findings.

So in case I haven’t convinced you, I offer one final reason why you shouldn’t use ChatGPT as an editor.

3. Even ChatGPT wants you to hire a HUMAN editor!

Because I’m a bit cheeky, I decided to ask ChatGPT whether or not I should use it to edit my academic writing. Here’s what it said:

Ok, ok. But seriously, don’t use ChatGPT to edit your academic writing. The lure of ChatGPT and other AI programs like Grammarly can be particularly strong when English is not a writer’s primary language. It’s even more important for these scholars to hire a human editor because ChatGPT is terrible at understanding colloquialisms and other nuances that EFL (English as a foreign language) speakers may also have trouble expressing or interpreting.

So, hopefully these explanations, examples, and ChatGPT’s own argument have convinced you that using AI to edit your writing is not a wise or effective decision. If you can afford to work with an editor, hire a human. We want to help you succeed, and we can promise—live and in-person (or at least over the phone)—that you won’t be disappointed.

Copyright 2025-2026 Paulina Cossette LLC – All rights reserved