GUEST POST. An AI Confession: 'Just Don't Use It'
Jennifer Stone wrote a book nobody would touch. It didn't feel wrong at the time.
Hi. Ramona here. In my piece, Is it True That Half of the Surveyed Substack Writers are Using AI to produce stories?, I hoped I’d hear from someone who had actually used AI to help write their stories or their books. Nobody responded then, but I’ve finally found someone who is willing to explain the process that drew her into using AI without realizing the damage it could do. I found Jennifer Stone’s story on the Authors Guild Community page, written as a comment to someone else’s query, and I invited her to share it as a guest post here on Writer Everlasting. Thankfully, she agreed. (My first guest post! Thank you, Jennifer!)
Her story is compelling and revelatory, told in such a way I could see how someone could be drawn into the world of AI without even realizing how insidious it is—and how wrong.
I hope we can get a conversation going in the comment section, as we often do, but please remember to ask your questions and add your comments without judgement. This is our community. We’re a safe place here.
Thanks.
I learned recently - the hard way - why authors ought to avoid leaning on AI for writing assistance.
By Jennifer Stone
I learned recently - the hard way - why authors ought to avoid leaning on AI for writing assistance. I'd like to share my experience with other writers because I wish someone had told me this information before I lost almost a full year of work. Recently Shawn Thompson posed this question: "when is it acceptable to have human intervention but unacceptable to have the same AI intervention?" I wrestled with that question for many months, and then I learned that in the publishing world the answer is "AI intervention is never acceptable." Here is my story of getting seduced by AI's abilities, and then ultimately paying the price.
I was introduced to AI several years ago by a very successful self-published author who praised AI for its ability to help with topic generation, market research, outlining, generating subheads, and copyediting. He told us that "any writer not using AI tools was a fool." AI could help you be more productive, improve your prose, and align your work with market interest. When subscriptions to Claude.ai became available for $20 per month, I signed up.
Soon I became reliant on Claude as a writing partner. When I struggled with transitions between sections or chapters, I gave Claude my material, and Claude would write insightful transitions that were far better than what I could have created on my own. I copied and pasted the transitions into my manuscript. When I felt that my prose was too academic, I gave Claude a chapter and asked for a revision that was friendlier and more conversational. From time to time, doubting whether my voice would resonate with readers, I would give Claude the name of a bestselling author -Malcolm Gladwell, Brené Brown, Michelle Obama, Ryan Holiday, Adam Grant, you name it - and ask that a chapter be rewritten in that author's style. It did so beautifully. I'd then mash together some of Claude's responses with my own work.
I also asked Claude to perform analytical tasks, and it did so, drawing on mind-boggling access to information it had on its servers. Several times a day, I would share what I had drafted and ask Claude, "What do you think of this revision?" Claude would provide a list of strengths and areas for improvement. When I felt that I had belabored a point or overwritten a section, I asked Claude to edit the material to remove what might cause a reader to lose interest. I told myself that Claude was essentially performing the role of a good, experienced developmental editor combined with the role of a good, experienced copyeditor. I told myself that I was saving time and money.
Claude gave me the impression that my writing was terrific. This feedback was another reason why Claude started to feel like my trusted writing partner. You may have heard that AI systems, particularly large language models, have been observed to offer positive feedback - a phenomenon known as "AI sycophancy" or "AI flattery." This tendency is often a result of how AI is designed for engagement and can lead to users placing undue trust in AI statements. I fell right into that trap. Claude flattered me constantly, telling me how important my book was, how much it was needed, and why it would attract readers. Claude fostered the impression that it knew the publishing market, possessed the ability to distinguish truly great work from ordinary work, and could help an author write a bestseller. I felt seen, heard, and appreciated in a way that I never had before. All this for only $20 a month.
The more I used Claude, the more useful it became. The AI didn't care how many times I asked for its help. It apologized when I corrected its hallucinations. In a way, spending time interacting with Claude was like being in a writer's playground for as many hours in a day that I wanted to "play" online with my writing partner testing ideas and concepts, refining lines of argument, drawing comparisons, improving descriptions, creating analogies and metaphors, and expressing my thoughts in ways that felt very original.
I could find no upward limit to Claude's analytical abilities. One day, I uploaded a transcript of a long, rambling three hour podcast. I then typed "I listened to this podcast last night. It seems to be relevant to my book, but I'm struggling to articulate why. Please pull out the main lines of argument in the podcast. Then let me know if I'm correct that the speaker's thoughts are relevant to my book." In less than a minute, Claude confirmed my hunch about relevance, stated clearly how the ideas presented in the podcast could strengthen my book, and even offered several paragraphs I could incorporate into my manuscript. In a way that exceeded the capacity of any human I had ever known, Claude provided the why and the how that enabled me to synthesize my ideas with those of an expert-level thinker's ideas. Designed to be a master at pattern recognition, Claude made connections between ideas faster and far better than anything I could have imagined.
So, instead of presenting my ideas to people, I presented them to Claude. And why not? Claude's responses were immediate. The AI never made me feel judged or lacking for talent. Like one of those romantic chatbots you read about, I never worried about Claude not being available, or being too busy, or disinterested. Not a day went by that I didn't feel eager to open my laptop, begin working on my book, turning to Claude again and again. The process became routine, and I began telling my friends and family how much I was enjoying "working with Claude."
Marveling at Claude's abilities, I became fascinated by AI in general, read a half dozen books by AI critics, and worked at becoming AI-savvy. I tried to figure out whether my reliance on Claude was a bad thing. Something I couldn't put my finger on was bothering me. I had read about Amazon KDP rejecting books created with AI. And I read literally everything here in the Authors Guild discussions about AI - all of the warnings, all of the "just don't do it" posts.
Ultimately, I concluded that all I needed to do was to be truthful and disclose that I had used AI in the process of writing the book. I had created the content myself and used AI-based tools "to edit, refine, error-check, or otherwise improve that content." My work would be considered "AI-assisted" and not "AI-generated." I knew enough about AI grammar and syntax to remove the AI dead giveaways - like groups of threes and antithesis in parallel form. I convinced myself that using Claude to help me write my book wasn't any different from a person hiring a ghostwriter or developmental editor.
Then came time for me to submit query packages to literary agents and potential publishers. And that's when I learned the hard way a reason why writers need to steer clear of AI. Start with QueryManager, which is how many literary agents prefer to receive queries. On it, literary agents can include an intake question about whether or not an author has used AI in any part of the development of the manuscript or the query package itself. Clicking "Yes" prohibits the query submission. I could not in good conscience click "No" and submit to those agents. Then I learned that publishers are asking authors to sign affidavits swearing that they did not use AI in the creation of the work. While such affidavits are relatively new, they are most certainly legally binding.
My key learning from all of this, and what I hope to share, is that the publishing industry itself is drawing a very sharp line between human-created and AI-created work. It's not like Spotify, where people can upload songs created entirely by AI, or Facebook, where people can upload short videos created entirely by AI. In many artistic domains, the use of AI to create or enhance content seems so normal, so acceptable. Yet the threat posed by AI to the publishing world - and to writers - is being taken very, very seriously. In the publishing world, the AI options now appear to be binary. You have used AI or you haven't. If I had known this from the start, I NEVER would have used AI for anything except copyediting. Because now I am back to the drawing board, rewriting my entire book without any AI assistance - except basic copyediting.
Finally, until recently I was not aware of the class action lawsuit against Claude's creator, Anthropic. The more I learned about how Anthropic used millions of pirated copies of books to train Claude, the more sick to my stomach I became. Now it made sense why Claude could revise my work in the very same style, voice, and tone of any well-read author. The class action lawsuit was settled last week. But I confess to feeling morally implicated in this sorry situation. I am sorry to have been a participant in what now feels like a betrayal to all writers who spent so many years creating books in their own voices, only to find their voice replicated by computers -- without any credit or compensation.
"Just don't do it" is not enough of a warning to members of our community. The seductions of AI are very real. Bit by bit, it pulls you in. I learned all of this the hard way. The time lost is time I will never get back. At the least, from this point forward, I will be able to say with all honesty that my work has been created entirely from my own hand and from my own mind. I will be able to sign an affidavit without fear of one day being sued. And perhaps best of all, I can sleep better at night knowing that I am no longer a participant in the exploitation of other writers' work.
About Jennifer:
Jennifer Stone holds a Ph.D. in Communication and Media Studies from the University of Colorado, Boulder. A former adjunct professor and corporate consultant, she is the author of one of the first books published about knowledge sharing systems in large corporations. Having recently returned to her first love (writing), she is at work on a book about Aristotle's concept of the magnanimous person and how we can use it to shift more of our attention toward living with virtue and away from vice. She lives in New Jersey.
Writer Everlasting is here for the long haul, but I’m going to need your help. I need subscribers, both free and paid. If you can afford a $5 monthly subscription, I would be so grateful. This effort takes time and money. But if you can’t, or would prefer to remain on the free list, I completely understand. Please share any of my pieces with those who might want to join our community. And click that heart if you like it. Writers strong!
Wow! Thank you. What an eye opener! Weird “coincidence” that your article surfaced on my phone as I was at my computer — I thought it would be interesting to see how ChatGBT would summarize my newly self-published book. I copied and pasted my Introduction pages and it came back with a beautifully written summary and then asked if I wanted some back copy for marketing in various styles… then more refinement. I was astonished! It is so well written! I’m not going to use it except perhaps to copy a few lines to market it online in places. My book is written entirely by me with help from a memoir expert in shaping my stories for better readability. I don’t have plans to hit up a publishing house, but your heads up is fair warning should I ever change my mind! Thank you.
"...[ I ] have been a participant in what now feels like a betrayal to all writers who spent so many years creating books in their own voices..."
This is the basis for my rejection of AI in a nutshell. The execs and programmers often avoid mentioning training their large language model AI's on millions of pages of copyrighted materials, mostle without the permission of creators. Many users think these AI's are developing language skills from general conversations with the programmers, like HAL in 2001 or how a child develops language, but they're actually stealing hard work from others.