"If the author isn’t responsible for the writing in their book, well, what are we even doing here?"
Preach. I can't imagine sending my nephew/boyfriend/guy-at-the-coffee-shop my novel then get an edited version back that I submitted unread to my readers or publisher. 😬
I've seen writers withdraw from contracts because the editorial hand was too heavy, but I've also seen writers who simply don't care and will accept any revision to get published and paid.
There are quite a few "Native American" writers who are indeed not Native at all but have fabricated personal histories. In Canada, these fakes suffer real repercussions. In the United States, these frauds really lose the jobs and accolades they received based on their lies. One of the most famous and recently-outed pretend Indians—pretendians—is Buffy Sainte-Marie.
He zeroes in on the fact that her writing method is the problem: She copy/pasted excerpts from texts she consulted into her own manuscript, not with the intention of integrating them wholesale into her own work, but with the idea of using them as placeholders she could consult and adapt as she worked on her own versions of those scenes. But this left her very vulnerable either to simply forgetting that those scenes hadn't been hers originally, or of struggling to come up with something sufficiently original once she'd grown used to those versions. I think a lot of college students end up committing plagiarism for similar reasons (i.e. an extremely flawed writing workflow).
I've been reflecting that I'd have a similar response to someone who said their work was, say, AI-assisted, or 10% AI-generated. Which 10%? I'd want to know. I think the not-knowing would take me out of the text. I'd want some way to gauge which parts were by an LLM, and I suspect that most authors using LLMs would struggle to answer that question by the time their manuscript was finished. (To anyone arguing that this would be a tricky question to answer even with two human authors, I say this: I would not really care about this if it was by two human authors, since the authors were human. Let the robots point to this quote in my trial for hate speech in 2053.)
In my teaching last term, I discussed something like this with my students & speculated that "best practices" in LLM use in the future might be a willingness to provide access to the full "conversation"/transcript of prompts and responses with the LLM as reference material for the project. I know some people are extremely protective of their prompts, but currently I have no real sympathy for that, no more than I have sympathy for being secretive about other reference material after you've completed a research project (apart from guarding vulnerable human sources, I guess).
PS: I hadn't heard about that Amy Griffin story. Lord have mercy, that's awful if true.
I wrote about many of these same items recently as well. What’s surprising to me is that publishers don’t do any research at all on memoir. I always had assumed they would do some minimal research. And today they are just becoming downright lazy, buying books that are selling without doing much diligence at all despite their names going on them. I suppose if the cost is lower than what they bring in, they’ll continue to take these risks.
If someone is being paid to do the thing, you expect them to do the thing - not steal from an existing book review, for example. So yeah, of course it matters.
Frey remains an interesting case study. The gatekeepers didn't like his books as fiction, which is what it was - auto fiction Repositioned as non fiction, whoo hoo, they were all over it. His AI effort was widely agreed to be rubbish, but at least he was honest about it, just as he gave it his best shot to be honest about his two books decades earlier. In some regards he was the canary in the coal mine.
Agree although I think it wasn't just the gatekeepers. I think the gatekeepers were correct in that readers would not have cared for it as a novel. At least, I cannot imagine A Million Little Pieces would have sold many copies if had come out as a novel. Certainly it wouldn't have gotten the Oprah book club treatment. The appeal seemed to be entirely with the rubber-necking of his (invented) struggles and then the redemptive memoir narrative. Obviously, can't know for certain. But even a lot of people who liked it as a memoir seemed to agree it was badly written!
Misery-lit sells. There remains a huge market, even on Substack.
I haven't read the book, I don't know how it would have landed as fiction, maybe it would have sold a respectable 50,000. It would have been a modest but honest sales number. That's not what the publishing industry wanted from him, nor, I assume, what he wanted for himself.
I think this is the key point with memoir - aside from the ethical issues (which remain important) it's clear that in many cases something which would have been an entirely underwhelming novel can seem legitimately sensational when sold to us as a "true story". It's exactly the same thing as a friend telling you an anecdote about something *that really happened to them* versus a story they just made up - why would we care about the latter??
What’s interesting is that ‘authenticity’ in writing has always been a feeling readers have, not a fact they can verify. The Frey scandal didn’t change because anyone proved the math on his fabrications — it changed when readers felt deceived. LLMs are the same problem, different mechanism.
Lately, two writers I follow on Substack admitted to using AI to write. They would tell AI the points, examples, stories, even jokes they have, and have AI generate a whole article for them. Then the writers would heavily edit the articles to check for errors, and make it sound like their voice. Yikes! I felt deceived and disappointed when I saw that.
One of them said that they're too busy with a full time job, pets and kids to take care of, plus they have ADHD, so they can't focus enough to write a whole article without the help of AI. They argue that the ideas, examples, stories, etc. are still their own, and they still heavily edit the article after. Sigh. I feel torn, because the points and examples in the articles are still very good, written by someone with the expertise to back it up too. (In fact, this person sounds like they care more about communicating information/ educating the audience rather than the craft of writing...)
I have ADHD too, so I take a very long time to get any article done. I'd rather publish slower than monthly (maybe bimonthly if I'm lucky) than to use AI to write, though. It still feels like cheating to me, even if the ideas, examples, jokes, are their own, and even if they edit it thoroughly after.
Maybe one issue is the belief that, if you want to do well on Substack, you "should" publish often, at least weekly. So for folks who want to succeed on Substack but are too busy/ can't focus enough to write that fast, they might take shortcuts to get there. :( Perhaps if we normalize slow publishing speeds on Substack, that you can still build a big audience even if you write at a snail's pace, then fewer people would be tempted to use AI to write their articles.
Yeah I was thinking of book authors (the topic of the linked article) BUT something like Substack is a place where it can be valuable to pump out content and quantity can matter more than quality, sadly.
Using AI to write a book review is really indefensible. Why would anyone do that unless they hadn’t read the book? Using AI to write phrases/sentences, whether in a novel or memoir, also seems indefensible to me—unless it’s to illustrate the nature/rapid advance/conundrum of AI.
I don’t know Lincoln everything I know about Large language models tells me you simply can’t get that kind of cut and paste alignment using a model. The guy must be lying. He may have pasted that this stuff in and asked the LLM to cover his tracks. Cut and paste is just fundamentally not how the models work. It’s a widespread misconception.
There’s been cases I know of when people cut and pasted material in from the Internet to their papers and meant to go back and rejigger everything to cover their tracks, but they forgot. Those people are known as some of my former writing students at the University of xxxxx. 😳
That might be right, but I have seen LLMs reproduce text and especially images—it’s quite easy to get them to recreate stills from famous movies—and with text it seems more common when there aren’t many sources to pull from. If this book didn’t have many reviews online when the NYT freelancer wrote it… But either way I definitely expect “AI did it!” to be a common excuse for various misdeeds going forward.
If this guy explicitly told the chatbot to base his review on existing reviews, and if there were very few or even one or two existing reviews, then his result might look very much like one of the existing reviews. Maybe that’s what you were saying.
Yes that's what I meant. When I ask ChatGPT to generate a review of that book, it does and seems to be pulling from the limited coverage of the book that's out there. It's not like asking it to discuss Machiavelli or someone who has extensive writing about them out there. There were a couple reviews and the publisher's blurb copy, basically.
The model is current to about Fall 2025 so when it concerns anything truly recent, there’s a high risk of theft here. All the lazy writers of the world better watch out!
I think you are right that it makes a person look less bad somehow to blame it on ChatGPT rather than just admitting they highlighted text and hit control C
"we can never prove exactly how much (a memoir is fabricated / a text used LLMs)" Oh I think we can. When AI copies song lyrics by Elvis and Taylor Swift, as I've pointed out recently, that's pretty definitive. Thanks Lincoln a good literary scandal is always a blast.
That certainly applies to plagiarism, but if James Frey says he spent three months in prison, who’s going to check court and other records to confirm that? Perhaps there comes a point where a pile-up of egregiously incredible details makes the reader doubt the whole and the more industrious investigate, though that seems a long way down the road for some memoir readers.
A book club I’m in favours memoirs so I’ve read a few recently, though it’s not necessarily a genre I’d choose. I always approach them cautiously and am unsurprised when, a few months after publication, someone pops us to challenge this or that detail. It happened with Tara Westover’s “Educated”, even though she explicitly included “other people remember this differently” caveats.
Thanks John, very accurate. Readers expect truth. When "egregiously incredible details" occur the reader can't trust the author, a bad thing to say the least.
Interestingly enough, the same is true for fiction. Good fiction has the ring of truth, and when details and motivations are unconvincing the reader's eyes glaze over. "But it really happened" is no defense. It's gotta ring true.
It occurred to me while reading this some of the uses of AI generally referenced are things I usually do with friends, namely spitballing and research. I feel like there's a point here to be made about this being another way AI might be replacing human connection, similar to how chat bots are being sold as lovers. Plus, I've often heard my husband (who is a historian,) complain about how poorly AI handles historical research. (For various reasons.) I'm no AI evangelist, nor do I think it's all pointless garbage. I take a similar middle ground stance. I'm just not sure anybody has really figured out what AI would really be useful for. I feel like it's going to end up being something really outside the box. :)
The idea of someone fact checking my memory sounds extremely uncomfortable. Nice point about our memories not never being 100% accurate. I remember reading about the majority of 9/11 memories being inaccurate - definitely made me question mine.
1.) I 100% agree on your stance. If an LLM did something for you that you would have to acknowledge if another source did that for you (editing, using as a main resource, quoting), then you should credit it. You may lose readers, but it's the honest thing to do. And honesty is typically rewarded.
2.) I don't know how James Frey is being used as an example of "well no one is criticizing him." I have never seen him not being criticized for something. Equating Mia Ballard with a known liar and thief, even before the AI stuff, is not the equivalence people are trying to go for. I don't get that defense for this behavior.
3.) Even if Mia Ballard didn't use AI, if she knowingly knew the book had AI (which it sounds like she did), I agree that she made herself look worse. What writer would let someone rework their work where it is practically rewritten? And if she knew, she should have disclosed it was edited heavily with AI. It's the only ethical thing to do.
I edit science manuscripts for a living and regularly spot plagiarism. For example, I found a sentence that was verbatim in at least 14 different papers by different authors. I flagged it because I had no way of knowing if the author self-plagiarized because he or she was anonymized. Before ChatGPT I’d have assumed they copied it and forgot they did so, but after LLMs it just makes me wonder how much of the actual manuscript is generated.
There's actually a novel written about a novelist who takes the life story of someone who has died and turns it into his own novel ... but then one of the people in the story pursues him for ... life plagiarism, or something. The novel is The Plot by Jean Hanff Korelitz.
An aspect of this that I find disturbing is that people only seem to care about the story of a life if they think it's true. What does that say about the readers? Why isn't a fictionalized life as interesting as a "real" one?
I must be the rare reader who prefers novels to memoirs, even if I suspect the novel is based on real life. The novel is likely to be better shaped and more imaginative than the memoir. It has more narrative drive.
Writing is hard. If someone's not willing to do the work themselves, they're not an artist.
Too many people want the status and/or money they believe to accrue to being a published author without having put in more work than having "an idea."
"If the author isn’t responsible for the writing in their book, well, what are we even doing here?"
Preach. I can't imagine sending my nephew/boyfriend/guy-at-the-coffee-shop my novel then get an edited version back that I submitted unread to my readers or publisher. 😬
I've seen writers withdraw from contracts because the editorial hand was too heavy, but I've also seen writers who simply don't care and will accept any revision to get published and paid.
Dear Sir or Madam, will you read my book?
It took me hours to write, will you take a look?
It's based on a novel by a man named Claude
And I need a job, so I want to be a paperback writer.
It's a thousand pages, give or take a few
I'll be writing more in an hour or two
I could make it longer if you like the style
I can change it 'round and I want to be a paperback writer
Sure. I charge $250/hour to read. Please send payment before you send pages.
There are quite a few "Native American" writers who are indeed not Native at all but have fabricated personal histories. In Canada, these fakes suffer real repercussions. In the United States, these frauds really lose the jobs and accolades they received based on their lies. One of the most famous and recently-outed pretend Indians—pretendians—is Buffy Sainte-Marie.
https://thehill.com/homenews/4314046-what-to-know-about-the-fake-identity-claims-surrounding-canadian-singer-buffy-sainte-marie/
When you say "these frauds really lose the jobs and accolades they received based on their lies," do you mean RARELY lose the jobs etc?
This made me think of a whole bunch of public cases of plagiarism in literature & how those went down. In particular, it reminded me of Jonathan Bailey's (I think pretty compassionate) response to the notice that Jumi Bello plagiarized his writing in her plagiarism apology (https://www.plagiarismtoday.com/2022/05/09/plagiarism-today-plagiarized-in-a-plagiarism-atonement-essay/).
He zeroes in on the fact that her writing method is the problem: She copy/pasted excerpts from texts she consulted into her own manuscript, not with the intention of integrating them wholesale into her own work, but with the idea of using them as placeholders she could consult and adapt as she worked on her own versions of those scenes. But this left her very vulnerable either to simply forgetting that those scenes hadn't been hers originally, or of struggling to come up with something sufficiently original once she'd grown used to those versions. I think a lot of college students end up committing plagiarism for similar reasons (i.e. an extremely flawed writing workflow).
I've been reflecting that I'd have a similar response to someone who said their work was, say, AI-assisted, or 10% AI-generated. Which 10%? I'd want to know. I think the not-knowing would take me out of the text. I'd want some way to gauge which parts were by an LLM, and I suspect that most authors using LLMs would struggle to answer that question by the time their manuscript was finished. (To anyone arguing that this would be a tricky question to answer even with two human authors, I say this: I would not really care about this if it was by two human authors, since the authors were human. Let the robots point to this quote in my trial for hate speech in 2053.)
In my teaching last term, I discussed something like this with my students & speculated that "best practices" in LLM use in the future might be a willingness to provide access to the full "conversation"/transcript of prompts and responses with the LLM as reference material for the project. I know some people are extremely protective of their prompts, but currently I have no real sympathy for that, no more than I have sympathy for being secretive about other reference material after you've completed a research project (apart from guarding vulnerable human sources, I guess).
PS: I hadn't heard about that Amy Griffin story. Lord have mercy, that's awful if true.
I wrote about many of these same items recently as well. What’s surprising to me is that publishers don’t do any research at all on memoir. I always had assumed they would do some minimal research. And today they are just becoming downright lazy, buying books that are selling without doing much diligence at all despite their names going on them. I suppose if the cost is lower than what they bring in, they’ll continue to take these risks.
If someone is being paid to do the thing, you expect them to do the thing - not steal from an existing book review, for example. So yeah, of course it matters.
Frey remains an interesting case study. The gatekeepers didn't like his books as fiction, which is what it was - auto fiction Repositioned as non fiction, whoo hoo, they were all over it. His AI effort was widely agreed to be rubbish, but at least he was honest about it, just as he gave it his best shot to be honest about his two books decades earlier. In some regards he was the canary in the coal mine.
Agree although I think it wasn't just the gatekeepers. I think the gatekeepers were correct in that readers would not have cared for it as a novel. At least, I cannot imagine A Million Little Pieces would have sold many copies if had come out as a novel. Certainly it wouldn't have gotten the Oprah book club treatment. The appeal seemed to be entirely with the rubber-necking of his (invented) struggles and then the redemptive memoir narrative. Obviously, can't know for certain. But even a lot of people who liked it as a memoir seemed to agree it was badly written!
Isn't that the point?
Misery-lit sells. There remains a huge market, even on Substack.
I haven't read the book, I don't know how it would have landed as fiction, maybe it would have sold a respectable 50,000. It would have been a modest but honest sales number. That's not what the publishing industry wanted from him, nor, I assume, what he wanted for himself.
I think this is the key point with memoir - aside from the ethical issues (which remain important) it's clear that in many cases something which would have been an entirely underwhelming novel can seem legitimately sensational when sold to us as a "true story". It's exactly the same thing as a friend telling you an anecdote about something *that really happened to them* versus a story they just made up - why would we care about the latter??
I'm not sure how well-known it is outside of the UK, but over here the recent Salt Path story is probably the closest thing we've had to the James Frey scandal: https://observer.co.uk/news/national/article/the-real-salt-path-how-the-couple-behind-a-bestseller-left-a-trail-of-debt-and-deceit
I love how Frey depicted Hazelden as a dangerous, gritty, lowlife dump. Yeah right. Celebrities like Liza Minnelli and Eric Clapton are alumni
What’s interesting is that ‘authenticity’ in writing has always been a feeling readers have, not a fact they can verify. The Frey scandal didn’t change because anyone proved the math on his fabrications — it changed when readers felt deceived. LLMs are the same problem, different mechanism.
Lately, two writers I follow on Substack admitted to using AI to write. They would tell AI the points, examples, stories, even jokes they have, and have AI generate a whole article for them. Then the writers would heavily edit the articles to check for errors, and make it sound like their voice. Yikes! I felt deceived and disappointed when I saw that.
One of them said that they're too busy with a full time job, pets and kids to take care of, plus they have ADHD, so they can't focus enough to write a whole article without the help of AI. They argue that the ideas, examples, stories, etc. are still their own, and they still heavily edit the article after. Sigh. I feel torn, because the points and examples in the articles are still very good, written by someone with the expertise to back it up too. (In fact, this person sounds like they care more about communicating information/ educating the audience rather than the craft of writing...)
I have ADHD too, so I take a very long time to get any article done. I'd rather publish slower than monthly (maybe bimonthly if I'm lucky) than to use AI to write, though. It still feels like cheating to me, even if the ideas, examples, jokes, are their own, and even if they edit it thoroughly after.
Maybe one issue is the belief that, if you want to do well on Substack, you "should" publish often, at least weekly. So for folks who want to succeed on Substack but are too busy/ can't focus enough to write that fast, they might take shortcuts to get there. :( Perhaps if we normalize slow publishing speeds on Substack, that you can still build a big audience even if you write at a snail's pace, then fewer people would be tempted to use AI to write their articles.
Yeah I was thinking of book authors (the topic of the linked article) BUT something like Substack is a place where it can be valuable to pump out content and quantity can matter more than quality, sadly.
Using AI to write a book review is really indefensible. Why would anyone do that unless they hadn’t read the book? Using AI to write phrases/sentences, whether in a novel or memoir, also seems indefensible to me—unless it’s to illustrate the nature/rapid advance/conundrum of AI.
Great article, thank you!
It’s strange to me because I read other reviews for the exact opposite reason: to make sure I don’t say the same things other people said.
I don’t know Lincoln everything I know about Large language models tells me you simply can’t get that kind of cut and paste alignment using a model. The guy must be lying. He may have pasted that this stuff in and asked the LLM to cover his tracks. Cut and paste is just fundamentally not how the models work. It’s a widespread misconception.
There’s been cases I know of when people cut and pasted material in from the Internet to their papers and meant to go back and rejigger everything to cover their tracks, but they forgot. Those people are known as some of my former writing students at the University of xxxxx. 😳
That might be right, but I have seen LLMs reproduce text and especially images—it’s quite easy to get them to recreate stills from famous movies—and with text it seems more common when there aren’t many sources to pull from. If this book didn’t have many reviews online when the NYT freelancer wrote it… But either way I definitely expect “AI did it!” to be a common excuse for various misdeeds going forward.
Correction…
If this guy explicitly told the chatbot to base his review on existing reviews, and if there were very few or even one or two existing reviews, then his result might look very much like one of the existing reviews. Maybe that’s what you were saying.
Yes that's what I meant. When I ask ChatGPT to generate a review of that book, it does and seems to be pulling from the limited coverage of the book that's out there. It's not like asking it to discuss Machiavelli or someone who has extensive writing about them out there. There were a couple reviews and the publisher's blurb copy, basically.
The model is current to about Fall 2025 so when it concerns anything truly recent, there’s a high risk of theft here. All the lazy writers of the world better watch out!
I think you are right that it makes a person look less bad somehow to blame it on ChatGPT rather than just admitting they highlighted text and hit control C
"we can never prove exactly how much (a memoir is fabricated / a text used LLMs)" Oh I think we can. When AI copies song lyrics by Elvis and Taylor Swift, as I've pointed out recently, that's pretty definitive. Thanks Lincoln a good literary scandal is always a blast.
That certainly applies to plagiarism, but if James Frey says he spent three months in prison, who’s going to check court and other records to confirm that? Perhaps there comes a point where a pile-up of egregiously incredible details makes the reader doubt the whole and the more industrious investigate, though that seems a long way down the road for some memoir readers.
A book club I’m in favours memoirs so I’ve read a few recently, though it’s not necessarily a genre I’d choose. I always approach them cautiously and am unsurprised when, a few months after publication, someone pops us to challenge this or that detail. It happened with Tara Westover’s “Educated”, even though she explicitly included “other people remember this differently” caveats.
Thanks John, very accurate. Readers expect truth. When "egregiously incredible details" occur the reader can't trust the author, a bad thing to say the least.
Interestingly enough, the same is true for fiction. Good fiction has the ring of truth, and when details and motivations are unconvincing the reader's eyes glaze over. "But it really happened" is no defense. It's gotta ring true.
It occurred to me while reading this some of the uses of AI generally referenced are things I usually do with friends, namely spitballing and research. I feel like there's a point here to be made about this being another way AI might be replacing human connection, similar to how chat bots are being sold as lovers. Plus, I've often heard my husband (who is a historian,) complain about how poorly AI handles historical research. (For various reasons.) I'm no AI evangelist, nor do I think it's all pointless garbage. I take a similar middle ground stance. I'm just not sure anybody has really figured out what AI would really be useful for. I feel like it's going to end up being something really outside the box. :)
The idea of someone fact checking my memory sounds extremely uncomfortable. Nice point about our memories not never being 100% accurate. I remember reading about the majority of 9/11 memories being inaccurate - definitely made me question mine.
1.) I 100% agree on your stance. If an LLM did something for you that you would have to acknowledge if another source did that for you (editing, using as a main resource, quoting), then you should credit it. You may lose readers, but it's the honest thing to do. And honesty is typically rewarded.
2.) I don't know how James Frey is being used as an example of "well no one is criticizing him." I have never seen him not being criticized for something. Equating Mia Ballard with a known liar and thief, even before the AI stuff, is not the equivalence people are trying to go for. I don't get that defense for this behavior.
3.) Even if Mia Ballard didn't use AI, if she knowingly knew the book had AI (which it sounds like she did), I agree that she made herself look worse. What writer would let someone rework their work where it is practically rewritten? And if she knew, she should have disclosed it was edited heavily with AI. It's the only ethical thing to do.
I edit science manuscripts for a living and regularly spot plagiarism. For example, I found a sentence that was verbatim in at least 14 different papers by different authors. I flagged it because I had no way of knowing if the author self-plagiarized because he or she was anonymized. Before ChatGPT I’d have assumed they copied it and forgot they did so, but after LLMs it just makes me wonder how much of the actual manuscript is generated.
There's actually a novel written about a novelist who takes the life story of someone who has died and turns it into his own novel ... but then one of the people in the story pursues him for ... life plagiarism, or something. The novel is The Plot by Jean Hanff Korelitz.
An aspect of this that I find disturbing is that people only seem to care about the story of a life if they think it's true. What does that say about the readers? Why isn't a fictionalized life as interesting as a "real" one?
I must be the rare reader who prefers novels to memoirs, even if I suspect the novel is based on real life. The novel is likely to be better shaped and more imaginative than the memoir. It has more narrative drive.
Agreed! I have no interest in memoirs. But novels: YES.