by Sheri Liguori, CopySoWrite – August 9, 2024 (updated March 26, 2025)

AI Gets a C as Editor

Following this overview of AI chatbots as collaborative writing tools is my list of AI’s Limitations and Gaps during editing. Copilot volunteered some of these points during a general Bing search.

ChatGPT vs. Gemini – and What About Claude?

The top contenders in the chatbot arena are ChatGPT, Gemini, and Claude. My initial practice was with Copilot (formerly Microsoft’s Bing Chat—accessed from the browser’s menu). From 2021 to the end of 2024, OpenAI’s ChatGPT Large Language Model (LLM) was integrated with Copilot. Since the end of 2024, Copilot has disengaged its alliance with OpenAI and subsists on its own. Though Copilot is built into Microsoft’s applications, I’ve conversed with it less, and log directly onto ChatGPT using my Google credentials..

For a real hoot, check out this article by Ryan Morrison on Tom’s Guide: I had Copilot and ChatGPT talk to each other—it got complicated.

Google’s Gemini is powered by the Gemini LLM. Gemini and Copilot are designed to integrate with your data.

Claude (created by Anthropic) was yet another LLM that rolled out in 2023 with free and paid versions. It has recently upgraded to Claude 3.7 Sonnet. I have conversed with it and find its responses to be very similar to ChatGPT’s. See one comparison on these two.

I asked each of these LLMs: Do you have the ability to edit a story? Claude’s response was the most elaborate.

Note: when first signing on, Claude says it “strives to do its best but is not perfect . . . may occasionally generate incorrect or misleading information, or produce offensive or biased content . . . not intended to give advice. Don’t rely on our conversation alone without doing your own independent research.” After each response, it also reminds us: “Claude can make mistakes. Please double-check responses.”

While writing this blog, a Gemini message popped up on my Android phone, inviting me to engage. I declined for now, as I don’t want AI butting in on my phone activities. I have played with Gemini on the Chrome browser only a little. For more on Gemini, a general comparison of it vs. Copilot and ChatGPT, and key differences between the pay versions, read this Popular Science article.

As of this writing, the Chatbot Arena LLM Leaderboard (user platform) ranks ChatGPT-4o #1, followed by Claude, then Gemini. Elon Musk’s Grok AI currently lags a fair distance behind. Originally, Grok had no free version, but as of February 2025, Grok 3 is free to all X users. There is still a paid "Super Grok" version for X Premium subscribers.

NOTE: Due to privacy and data protection concerns, I have not evaluated China’s DeepSeek AI (it launched for public use in January of 2025).

SLM Apps Outshine Word’s Spelling & Grammar Check

Small Language Models (SLMs) such as Grammarly, Scribbr, Wordtune, Hemingway, and others can be helpful, but understand this: they don’t catch every problem and often make inappropriate or incorrect suggestions, introducing new errors. This has been my experience and is a consensus among editors. I’ve had fun playing with the free versions of SLM apps and find it interesting when one arrives at slightly different corrective advice than another.

Microsoft Word’s grammar check function is not particularly robust, but it finds most misspelled words. Despite its limitations, Word remains the preferred writing canvas for editors working with authors due to its Track Changes function, which allows for easy tracking of edits and comments.

The Invisible AI Writer Unmasked

According to Ethan Mollick in his book Co-Intelligence: Living and Working with AI published in 2024, “there is no way to detect whether or not a piece of text is AI-generated.” Don’t bother using an AI detector app because none of them are foolproof, and often result in false positives (pasting text from grammar apps into your work could inadvertently introduce embedded code, also potentially triggering false AI detection). Still, there are ways that editors and very astute readers can tell when AI has done some of the writing. This article “13 Ways to Detect AI Content” by Satya D (SEO specialist and Surferseo.com blog contributor) offers some insight.

What happened during pretraining or machine learning? AI was fed data from the open web and books (by computer scientists and underpaid entry-level folks), which explains how biases and misinformation could not be completely filtered out. Apparently, no authors were contacted to obtain their permission. Some publishers have since openly allowed LLMs to be trained on their books, whereas other publishers are careful to acquire signed contract clauses concerning the use of such content for LLM training—or not—with authors and agents.

AI as Editor: Limitations and Gaps

Editors’ Perspectives on AI and Editing:

Wes Cowley, a fellow freelance editor who spent many years in the IT industry writing and editing technical and business material, is firm about his position never to use LLMs in his writing or editing work. His reasons for this are fully explained on his website.

But wait, there’s more! Rachel Baron reported on AsiaEdit’s extensive testing of generative AI tools for editing work in a recent abstract article for Science Editor. For a simplified version of the main points, read Wes Cowley‘s summary.

Learning and Earning With AI

While AI may not be the perfect editor, its role in writing and learning is evolving rapidly, with implications for both education and the workplace.

Mollick, a Wharton professor specializing in management and innovation, actively incorporates AI tools into student learning. In his aforementioned book (pp. 94–96), he delves into the underlying reasons why LLMs are prone to hallucinate. That said, while some educators ban the use of AI in their students’ coursework, Mollick requires that his students use AI for various assignments. He explains that today, students are already using AI as a learning tool, and many teachers have begun to use AI in preparing for class.

On the future of AI in the classroom, Mollick says: “We’ll find a practical consensus that will allow AI to be integrated into the learning process without compromising the development of critical skills. Just as calculators did not replace the need for learning math, AI will not replace the need for learning to write and think critically . . . we must do so—it’s too late to put the genie back in the bottle.”

Even the free online education platform Khan Academy uses AI tutoring. A 7:20 video explains how chatbots and LLMs work. For the latest in classroom AI-assisted learning, watch this CBS 60 Minutes report on how Khanmigo, an AI-powered learning tool developed by Khan Academy in partnership with OpenAI, is being tested in schools.

Mollick’s research on this topic as explained in his books (some of which was performed in partnership with lead researchers from other universities) is fascinating and easy to grasp for non-tech types. He is certain that learning to work in tandem with LLMs will ultimately be beneficial, and I would add: especially with continued standards for AI safety and security. “Jobs don’t consist of one automatable task, but rather a set of complex tasks that still require human judgment,” he says. The world of work is headed in this direction, whether we like it or not.

Looking for an editor? If you’re interested in working with me, Complete my form or email me about your project. I will contact you to discuss your needs.

The Creation of My Logo:

Additional recommended reading:

Why great AI produces lazy humans, Ethan Mollick, Big Think, March 26, 2024.

How to succeed at the “jagged technological frontier” of AI, Kevin Dickinson, October 23, 2023.

‘Worse’ AI Counterintuitively Enhances Human Decision Making and Performance, Adam Hadhazy, July 5, 2022.