Why don’t we start requiring authors to highlight which lines of dialogue are theirs and which ones were modified by a human editor? We don’t do that because it’s fucking absurd.
so how is hiring a ghostwriter any different than using an AI to produce a book? in terms of ethics. they do not always disclose if a ghostwriter was used. the only way you know is if they produce so many books in a short period of time, or if there is a "with" credit on the cover or "as told to" type of thing.
ghost writing has been happening for at least 100 years with books. HP Lovecraft ghostwrote a story. The Hardy Boys and Nancy Drew series were all ghostwritten. There are modern book series being ghostwritten which are obvious when they seem to be able to release a new novel every month and they've been doing it since before AI was a thing. I doubt Blake Pierce is writing all their own novels, for example
Exactly. If one wanted to be outraged about anything, it should be that. Already-rich celebrities hitting the best-seller list for something they didn't write a word of.
At the end of the day AI a tool originally designed to help is being abused by the rich and powerful to keep the 1% on top and marginalize everyone else: they will always try to silence our voices, spreading the word is how we help each other speak out
But it's not working that way. Ordinary people ARE using AI for all kinds of helpful things. There is nothing wrong with that. AND, it is not illegal, nor is it anyone else's business. If they want to do that, let them. If you don't want to do that, then don't.
Truthfully, yes, AI is very helpful for a lot of people and however comfortable you are with using AI or not at all is entirely up to the individual. Sadly, like many societal issues where privacy should be paramount, “everyone” has an opinion and has to make sure “everyone else” knows it, even if they have no factual support or research, just rumor and conjecture. However, when patterns start to emerge like AI claiming that several successful writers from the same marginalized group where English is not their native language and flagging them to the point is costing successful and talented writers publishing deals, that’s not a simple fix, it’s a recurring problem that needs to be addressed before it becomes an even worse problem that xoukd lead to some really bad conflicts
I myself will no longer be reading self published works published after 2020 (thought it might go to 2018) because authors are simply not honest about their use of ai, and because the use of ai goes far beyond simply whether its your voice, far beyond the fact that its theft (in fact I'll probably be blocking this blog because it uses ai images, but I respect the writer for disclosing so I can make that choice), and far beyond the environmental impacts.
I wish those that champion it would do a much deeper dive on the ways it is harming so many people and the future and stability of civilization.
I'm sure others are not far behind in ditching books written during a certain time period because the trust is simply not there. So feel free to tell the ai police to pound sand and self publish, but I recommend when you do, you at least be as honest with your readers, as the author implores.
In the context of raging AI debates, I offer this Substack PSA:
<SUPPORT YOUR LOCAL POET> I self-published two books of poetry--poetry I wrote, read aloud at open mics, edited, rewrote, shared with other writers, revised, performed, collected, and proofread. A book designer created and uploaded the cover and contents files. Printed on demand, the books are beautiful and still selling, one here, one there. If you would like to purchase a non-AI collection of smart, accessible, well-written, meaningful poetry--not the kind you have to decipher, but the kind you read, think about, and read again--DM me for a signed copy. Or find *WHEN EAST WAS NORTH* and *RABBIT SUN, LOTUS MOON* on Amazon, by me, Andrea Millenson Penner. Or find my 2024 chapbook, *Upcycled: Poetry Repurposed* new and updated work because "like people, poems deserve a second chance." Published as Andi Penner. Thank you for supporting human-generated creativity. Sent with love and gratitude.
I don't care if someone has used AI or not. I doubt I've ever read a GenAI novel as I nearly exclusively read works from the 20th century and earlier. But if I'm reading an article and suspect, or later find out, that it's AI - so be it. Was it informative? Was it engaging? Was it clear? Did it seem accurate and truthful? That's all that matters to me.
This right here is why I hate GenAI. I began working on my debut novel in 2015. I paused to get an art degree so I could retain as much creative control over my novels as possible by doing all my own covers and marketing for self-publishing. I do not use AI. I roll back updates and disable every AI "tool" every time I can. But it doesn't matter that I have all the receipts to prove I wrote my novels word-by-word. I can't go back in time to publish it before 2020, so it's assumed I used AI and am lying about it. I hate this timeline!
Social media has been doing (in many regards) enormous harm to people and civilisation for nearly 20 years now but that hasn’t stopped millions from using it all day long.
Wow! Thank you for putting so succinctly what I have been thinking. Adding to the issues you mentioned, it's the tone of the article or a piece that is annoyingly similar to other AI generated piece. It feels like waking up to someone's shrill voice and hearing that all through the day, over and over again!
Gordon Lish and his overbearing editorial control of Raymond Carver's entire body of work is now considered problematic, but that took decades. His books are still available.
He's quite famous for his short story collections. Lish's career is equally infamous.
"Lish's cuts were sweeping, wholesale, and so violent that Carver described them as "surgical amputation and transplant". He responded with a now-famous cri de coeur, writing a letter to Lish explaining his horror: "I don't want to sound melodramatic," he wrote, "but I've come back from the grave here to start writing stories once more . . . I'll tell you the truth, my very sanity is on the line.""
I absolutely adore Raymond Carver. I spent a long period of time trying to emulate his writing with my own short stories. Nothing as obvious as plagiarism, but someone that I greatly admired.
Yes, I love his writing, or what was left of it. A unique voice, difficult to emulate. Writing style and subject choices for fiction are much like finger prints.
I was gutted when the Lish controversy became public, more gutted for Carver than myself as a reader. I wondered why Lish didn't stick to his own writing and spare other authors his editing.
Imagine having an editor who changes the entire intent of your writing.
It is the very reason I gave up on human editors. This is impossible to explain to people that have already decided that there is only one way to leverage AI when writing, and that is copy and paste, which is absurd.
In the same way that one can ignore a human editors suggestion you are free to ignore the AIs editorial suggestions as well, it is really that simple. That is how you keep your voice.
Right, and the post above actually gives a way to do that without turning it into a disclosure fight: a per-session log of what you asked for, what you took, and what you rejected. Nobody publishes that log, it just sits with you. And the rejected column ends up being the interesting one over time, it is not just a record of edits you skipped, it is a pattern of exactly where your judgment and the model's default keep splitting, which tells you more about your own voice than any single accepted line does.
Thank you. I wonder why AI is so bad? I use it frequently in my writing. I feed it information, it gives me sentences and paragraphs. I accept or reject and request a rewrite. I feel like it is only a tool to help my work better and faster. If a book is good, I hardly think AI wrote it from a single prompt!
They frequently do, which is how authors get caught out. The same with commissioned custom art, particularly when it's cheap. That said, I don't see how a GenAI artwork cover is any worse than using the same stock photo as thousands of other authors have, which happens all the time in genres like Romance. I've bought stock photos for covers before and I've seen at least one of them used by another author. You can buy exclusive licences from some sites, where they then remove them from further sale, but why should we? As a reader I couldn't care less about it.
I completely agree. I think the issue is licensed resale/reuse versus nonsanctioned edits of an original work without the authors acknowledgement or consent AND/or AI used under the auspices of original work. It changes the landscape of creative and original work. I’m not against AI, either. I can just understand the argument on both sides.
I am so torn about this topic. On the one hand, I‘m a literary translator and will likely lose my job because publishers and authors are unwilling to pay me a living wage anymore, and I detest dishonesty and plagiarism. On the other hand, I‘ve read many amazing stories in the realm of amateur fiction that would never be picked up by a publishing house and for successful self-publishing would have required the work of an editor who these gifted writers could never afford. So like the options for self-publishing I see this AI as an important step towards making publishing more equally accessible. I‘ve also encountered a creator who had AI write his complete novel. He was so generous to share his complete, painstaking process from building the plot and the characters to perfecting the finest nuances of voice and style. No, he didn‘t write the novel, but he… was the architect responsible for the placement of every word. The product was a fascinating read that shouldn‘t be dismissed as a literary creation. I‘m just sad for the creator that he missed out on the experience of writing and authorship.
In terms of getting published, I must admit that I am deeply discouraged. Even before the advent of AI the way authors increasingly had to put in massive efforts in terms of social media and how "shitstorms" had basically become a given and pretty much unavoidable made the idea of publishing less attractive. And now this…
Your experience provides lots of food for thought. It's definitely not a completely black or white situation; there's plenty of room for multiple grey shades in the middle. At the end a tool is just a tool and it's in the way we use it, like fire, like a knife, like the wheel. Steph, I too am not sure what the way forward is, but what scares me is not artificial intelligence. It's human idiocy, and our proven tendency throughout our many civilizations to self-destruct, instead of learning to live harmoniously and cooperatively.
I definitely agree with you there human stupidity is pretty terrifying. What’s even more terrifying to me is that I fear that this is just how we are. This is what we are as beings and that we won’t change. Maybe we can’t change and if that’s the case that’s sad.
I have the outline, only, of a book I wish to write, based upon personal experiences. I do not intend to use any AI, or grammarly at all. I do intend to use a human editor. I have zero interest in these tools. Thank you for this information. I will be saving my paper trail.
Of course, no. I was edited once, by my own mother. In junior high, she typed up a short story I had written for class. She changed a sentence I had written from “Wait for me!” to “Wait! Oh wait for me!” It was too late to change it back. My typing would have taken another two days, with eight false starts. I tell you this to say her editing one word without me agreeing to it is still under my skin after 50 years. So it is unlikely anyone will be significantly editing anything other than basic grammar. I will take suggestions, but it will require discipline to avoid acting on that same childish feeling when it wells up. Sad when emotional responses fail to evolve.
Nah because you can - because that's called a saved copy and tracked changes, comments turned on.
Ready to prove my proof of process… we have been ready since we learnt ai has stolen our text… with enough to take multimillion dollar companies to court.
Also, that means a court date, coz I shall be compensated for my time :)
I'd be interested in a statement from the publisher as to their objection to minor use of AI. If their stance is that it's not the author''s own work, I don't see how that is defensible. Using Grammarly means it's not all the author's own work. Ditto using a copy editor. Ditto asking friends to read a draft and give feedback. And using a ghost writer?
Maybe it's time for Arthur Miller to rewrite The Crucible, replacing witchcraft with AI.
This is exactly how I feel as well, Martin. Similarly, for the extent to which tools are used for editing, even in Microsoft Word, we've genuinely got to a point of insanity at this point, and you're absolutely right to point to Miller's work is a very depressing touch point.
I can see typewriters coming back. Think of all those books pre digital written by hand and edited by hand. Authors have become lazy and AI is the ultimate laziness. Challenge is where AI is used to edit (that is tweaking) versus it being used to write chapters that are then edited by humans to make it less AI written. Maybe we will see disclaimers in books ... this book was drafted on paper with an ink pen and here is a video of them at 3 in the morning writing chapter 6 etc. I want my authors to have struggled in their writing ... thats what makes them authentic but it seems from the research most readers dont care as long as its a good read
Do you think authors who use ai don’t struggle? You are mistaken. Why is anyone writing their replies here on social media on the internet? I propose henceforth we go offline and correspond only by handwritten notes delivered on horseback (only by men because clearly women don’t belong on horseback).
Martin's right that nobody tracks Grammarly or a good copyeditor the same way. The distinction the post's own fix points at isn't purity, it's disclosure. An audit trail log doesn't ask whether a tool touched the work, it asks what you asked for and what you kept versus threw out.
That's a different question than "did AI touch this," and it's the one a publisher can actually check, since the log just shows who made the final calls on the page.
But then a copy editor is another professional who is doing a job and getting paid for it and contributing the publishing industry. Using AI removes that cog and destabilises an already fragile and vulnerable system in my opinion (but I’m not an academic, and have never published a book - just an enthusiastic reader who’s watched my own industry crash and burn thanks to AI)
And writers likely pay to use AI. Virtually impossible to write a book and keep track of everything in free versions. And any AI would lose track of characters, story arc, etc. without human intervention (at least for now). It struggles to write a cohesive chapter without guidance.
How editors work is they suggest changes. They do not make changes - i think even a semicolon instead of a period would have a comment as to why, and that is with tracked changes on even if they did just go ahead and fix a comma placement. (One would hope to do their own line editing before it gets to a human editor who charges by the word. Imagine an editor having to fix your commas routinely.)
Anything that inserts words or changed them around is not being an editor, it is doing the writing…
Which is required to be disclosed.
A dumb spelling and grammar checker like in windows 95, is a tool suggesting the change of affect to effect. Links, the grammar assistant cat in windows 95 has a pop up bubble that says “did you mean effect?” And gives a little explanation as to why and the definition.
I haven't a clue about grammarly, but nothing and no one should be changing words around. It, or the human editor, should only offer suggestions for the author to decide to implement or not. I think a lot of people do not know what editors do and what they don't do.
That's not quite right about editing. When I edited journals, I changed sentences around if they were eg. from European authors who had used European sentence structure. In fact I edited everything that needed editing to make it sound like UK English.
The onky thing I checked with authors was things in equations such as if an *x* needed to be italicised.
Interesting! I never had an editor move things around for me or change my sentence structures at all. Comments only, then comma corrections. When I did a cert 4 in professional writing and editing with Tafe gippsland and then diploma with university of new england... We were taught to be very non intrusive, as it wasn't our work. 🤔 Interesting.
Yesterday, i cut and pasted the first page of a celebrated 1974 book called The Dispossessed by noted author Ursula Le Guin into an AI and asked if it was AI assisted. I did this because I was shocked by the first two lines and each subsequent line because that page contained ALL of the major AI tells. Line after line after line. The LLMs borrowed heavily from her prose. The AI came back with 75-85% AI and a few pages of all of the justification given in various articles about contest winning and publication receiving authors outed as using AI. But of course this work is not a product of AI. 🤖
So the real objection to AI seems to be that AI is copying the style of established (or any and all writers) and offering that style to anyone using the AI. True?
Correct. The first page of that book is the most perfect 1:1 representation of what LLMs copied from an author to structure its output. The tail is now wagging the dog.
I will never pander to bad-faith accusations of AI use, and I will never keep an audit trail. The work *is* the audit. How it got onto the page is no one's business. Whether a suspicion machine like Pangram likes it or not is irrelevant. AI is a research/writing tool and a universal translator for human ideas. It was trained on the best of human writing. But raw model output is generally unsuitable for use without extensive human editing anyway. Which is how it should be. Every author must stand behind their work regardless of their process.
Eventually this controversy will die a well-deserved death and people can get back to judging writing on craft, along with quality of thought and argument.
I feel the exact same way. This nonsense is all the fault of the anti-AI crowd who care less about content than they do about sticking their noses in someone else's business. It shouldn't matter how the words got on the page if you enjoy reading them. I've never once asked myself about a favorite author, hmm i wonder if all this writing is theirs or has it been extensively rewritten by editors. Because I don't care. I only care that I enjoy reading the book.
I totally agree. How is this different than that of an editor, copywriter, or ghost writer? When I buy a book, I'm not doing an over-arching detailed analysis of the words on the page. Honestly, I don't care. I will pick up the occasional typo because secretly I'm a grammar-nazi, but whether a turn of phrase is created by the author, editor, or AI - doesn't concern me.
I'm anti-mimmix (AI), but don't report AI slop. It's a useless endeavor, as that's part of training mimmix & removing human authorship from the household. I refuse to be part of either end of that.
You don't? You don't think about the moral lessons in a story? Or the real life stories that ground what has been written on the page? You don't wonder what an author meant by something? Or if it was changed or modified by an editor? You don't find joy in seeking out what they as a person was attempting to say? You think only of its purely entertainment value? You don't wonder about your favorite authors as people and how that impacts what you're reading?
Lol it's really not. People who don't write fiction think it is because they believe their high school Lit teacher who designed entire classes on figuring out what an author meant by something. It's actually pretty funny. All the patterns in my fiction put themselves there without my conscious thought. But lol non-writers think we're constantly threading in our own values and lives into our fiction and making esoteric points through metaphors or whatever. Maybe some do, but I'm not one of them, and i doubt I'm alone. The things I write about are very far removed from my actual life or experiences. That's why it's called FICTION and specifically, speculative fiction. Imagination means you can imagine lives that aren't your own.
This is actually a fascinating claim. You believe all your writing has been written without conscious thought. That lessons, habits and biases of your own life have been left out. While you sit down to write who do you envision is writing on your behalf? Is this same source writing for all others? Is it only in fiction where this occurs?
The reason I keep an automated audit trail is so I can roll back misguided changes and keep track of why I made certain decisions. That just happens to be an excellent audit trail in general.
The AI police are constantly talking about how bad AI writing is but they rarely define what “ai written” means. Is it okay to use it to edit? Is the problem generating original content with the tool? They don’t know. All they say is that AI writing is slop. But if it is so bad, it should stand out like a sore thumb. And it frankly doesn’t in most cases. Because in most cases authors who are using it are using it as a tool. They are taking some of what it gives, rejecting other stuff, and heavily editing anything it produces. I completely agree, let the prose stand for itself. It’s not my business or anyone else’s how it got there.
Publishers are doing this to avoid lawsuits. They don’t want to get accused of copyright infringement. I understand protecting themselves from a business standpoint, but they’d do better to define their terms up front and add an AI indemnity clause to the contract.
Anything creative requires tools. If the creator using the tool is mindful, and allows common sense and wisdom to guide their decisions in crafting a work — whether it is a brush on canvas, hands on a drum, or fingers on a keyboard — what interests me is the final piece. Goodness knows I spend just as much time babysitting Claude and correcting it than I do crafting what I write so the written word remains true to my voice. I care about the craft and interior process of thought, memory and inspiration deeply. I am also immensely grateful for the edits and audits available to me, a true gift as a pen-mute woman who disappeared from life for seven years and needed a way back to words. I never could have imagined what form that support would take.
I mean it matters because writing isn't judged primarily off of craft but off of what one person is attempting to say to another. When we add more and more things between those two people (like AI) we're very naturally going to run into issues…
I am not doing any of this BS. According to my stats no one really reads my work anyway. This seems like double the work for the same amount of non-readers and non-fans. My advice, just go read something else. I honestly do not care.
I absolutely respect your right to say this, Kenneth, and I genuinely believe that within 12 to 18 months we'll be entering a phase where a lot of this will be by the way anyway. At the end of the day, all AI really is is a selection of tools that, when harnessed correctly, can massively improve the quality of writing.
You are in good and plentiful company when you say AI "can massively improve the quality of writing."
I, however, must have a different definition of "quality." If the grammar is perfect but the writing is now bland, lacking any enjoyable voice or personality, I would not say that the quality is improved.
AI writing that's so obviously AI, no need for a detector, is mind-numbingly boring, and most importantly, it's really sad. I know writers whose content is interesting, and they say "I write it and then ask AI to rearrange and tighten it up." What they're actually doing is using AI to strip out any interesting parts of themselves from the writing, their own human quirks and thoughts, in order to make it "better."
That is a loss, for both writer and reader. That is why I wish writers wouldn't use AI. I will never reread their books the way I reread the authors whose voice is delightful, or fascinating, some books so many dozens of times that the binding falls apart. But that is how I define "quality" in writing. And it is something AI will never be able to do.
This is a great question, for me, the difference is that an AI editor does not decide what my motivations must have been before reading what I actually wrote. Using AI as an editor helps me focus more on narrative structure and refinement, not managing the personal biases and ideological imperatives of a human editor. I don’t have to justify what I choose to write about.
AI editors do not turn every editing session into a long back and forth lecture about what a “Male” or “Black male” writer “should think” and write about when it comes to issues of race, gender, or politics. I don’t need lectures about “the patriarchy” and “male privilege.” I don’t need lectures about what I do or don’t have the right to talk about because of who or what I am.
When I put hands to the keyboard THAT is what gives me the right, I have the right because I take it! I don’t need anyone’s permission to create a character or address a topic of my choosing. What I want from an editor is actual engagement with my work, and that is what I get from using AI as an editor. Example;
Where can pacing be improved?
Does this scene make sense?
Does this characters argument make sense?
Is the prose clear?
Is this order of actions in this scene clear?
What can be improved in this dialogue exchange?
Are the character’s actions or reactions believable given the way this scene is set up?
Where am I repeating myself?
Have I failed to explain something clearly?
In my view this is the advantage of using an AI as an editor over tolerating a human one.
Why is it so easy to tell when an AI edited something? That's the essence of the difference between human editors and AI.
It is heavy-handed in specific ways-- it pulls all writing in line with a generic standard on specific lines- and by the way, pulling it up to standard, or down to standard- and makes it all sound vaguely the same.
It doesn't know any better. It can't. It's programmed with an as-yet-limited hierarchy of "these are best- good- acceptable ways of doing this" and that's the model to which it sets any writing.
Maybe this will become less obvious over time, but until AI has a soul, it won't be able to have an author's voice.
Human editors, in general, do not do this to such a broad extent. Not to claim structure is never addressed by humans, but they don't make everyone's writing sound the same, with the same structure and same generic catch-phrases. Because it's a human, looking at your work with their human brain and making it better, not indiscriminately molding it to a predetermined generic standard.
I loved this comment. And I agree. I actually had to be very stern with the AI I finally chose to help with my writing and make it clear that in no way shape or form is it to ever even suggest an edit that would follow modern writing standards of tailoring to the weakest literacy link or strip my writing to read like boring corporate internal memos. My audience is not children's lit and my literary role models that I try to emulate are King, Koontz, Twain, Hemingway, Browning, etc. Literacy of that caliber. I may not be as great as those writers, but they are the masters at storytelling. I refuse to let the AI try to strip that from my writing for the sake of making it easier to read.
I have even contemplated pulling my fantasy from the one website where I share it simply because my readership fell off the minute I took down my first draft and posted up my second draft. The readers seamed to prefer the stripped down, less immersive version I expanded it from. I often wonder what happened to all the readers like me who love being able to feel fully immersed and lost in the story.
That is all well and find but again, there is not just one way to use this technology. People that assume that are just misinformed. Some of my earliest posts explain multiple ways writers can use this technology. (see below) I think I have covered precisely how I use AI pretty well, so I really feel no obligation to explain it even more.
The reality is that I gave up on human editors years ago. Once you sit through enough finger-wagging lectures by ideologically driven editors, you reach a point where enough is enough. I have found that human editors assume your background and writing motivations based on their biases. This makes the ease and practicality of using AI for feedback, editing, proofreading, and story structure checks much more appealing.
If anyone has a problem with that, or my work the block feature is at their disposal.
I write something, ask AI for critique, then discard half to two thirds of its suggestions because, yes, they would sand off the edges and make it no longer sound like me. But the third to a half that I consider and then implement my own way do I prove the work. It's a tool. And like any tool, requires discernment in use.
People that have nothing better to do worry about it. Then again, I’m sure they look at every photo with tears in their eyes wondering whether or not Photoshop was used on that photo before they get to enjoy it.
It really is a witch hunt which is precisely why we need to do what people during the witch trials should’ve done and that’s to tell these people to #uck off!
The sad part is a lot of people are just going to be lousy on purpose because they’re afraid that if they write the way that they normally do (with many of those telltale signs of AI which quite frankly are just part of human writing) they’ll get flagged and shamed for using AI.
Why don’t we take a lesson from the witch trials and simply tell these people to fuck off instead of making complicated statements about how we did NOT fornicate with the Devil?
To hell with the how I made this statements. To hell with the auditing trail for people who are strangely more concerned about tool usage than the content of a persons work.
I am not particularly political, and I have already written about ways in which writers can use AI. Some of my earliest posts talked about this topic and I think I have covered precisely how I use AI pretty well, so I really feel no obligation to explain it even more to people that would otherwise not even read my work anyway.
Again, if anyone has a problem with that, the block feature is at their disposal.
Looks like I was caught being autistic in public again. I meant no offense. I actually agree with you, and I was just pointing out that if people want to shame me for using AI, good luck with that because I'm not ashamed. I don't buy into their frame about any AI use tainting the entire work. As long as I write all the words myself, which I'm going to do because that's the fun part, they can go to hell.
The audit trail idea is practical, but it’s also a little heartbreaking that writers now need a paper trail to prove they created something. It makes me want to stop writing and go back to journaling :(
Does any serious writer write a poem let alone an entire novel without saving versions, research, notes, edits? There would always be an entire cache of material. I can't imagine anyone blithely overwriting every version of their drafts, or tossing away research once used. Even in business, people save versions.
There's so much wrong with this lastest episode, too much to go into. But at the top would be the lack of due diligence PRIOR to selling a manuscript, and if LLMs can churn out a novel so compelling that it's worth $2M up front, wouldn't we all be doing it? We're not, because LLMs can't produce any fiction that good.
I don't keep drafts of my writing when I rewrite. The clutter bothers me and it's like building on a existing foundation so I have the old memorized and only needing the new changes. Although most of my world building is in my head and done verbally when explaining my story to someone. I guess the main audit trail I would have is that I listen my different iterations so I have a lot of elevenreader files. Although for me I'll tell them here's my writing I want to see you FORCE AI to write like me because I know it can't.
Honestly for most your statement would be perfectly accurate, but I feel as if the desire to publish and having to confine your writing to a box to fit through the gate has made people forget that's it's Art. That's it's creative and free like water and from person to person it's all different and SHOULD BE that's what makes it amazing. If we all follow the same process and patterns then we really aren't much different from AI in my opinion.
😂 Overall I've learned my mind works oddly so while I agree with your answer I'm always worried about the outliners .
I am careful with document control at work as a public servant. But I confess I do a lot of overwriting for my Substack series. I've only recently started seriously considering turning my work into a book down the track. So I appreciate the advice.
Ah, yep, all too familiar with document management in the public service!
I definitely don't miss the torture of dueling edit tracking. 😂
I don't use edit tracking on my own work, I would find that weird, but I do create an electronic filing structure, and save versions under individual chapter files, or research topics, etc, using the normal v.xxx format and a date as part of my naming convention.
I agree with you, Caz. And if I ever finish the book I've been threatening to write since 2002, there will be plenty of evidence. 😂
But I was coming at this from the perspective of a nobody writer on Substack (me personally).
I'm thinking defensively. What happens when an ordinary writer gets accused of using AI?
That's the part I find depressing.
I also agree that the publishing world needs to check itself because I genuinely believe this author was wronged on so many levels. The problem is his name has already been tarnished.
Agree, and this writer was a first time writer (as was Shy Girl ?), so an ordinary writer, if you like, until it all went pear-shaped with a $2M deal. From the little I've seen, the agents withdrew the manuscript, it sounds like he had no audit trail to share with them (why didn't they ask before putting it up for auction?) He first drafted the book several years ago, apparently subpar. The next step was a great manuscript. Nothing in between.
I would hyperventilate if I didn't have every version. I know that's not how everyone works. Oddly, I think people really putting in the work, bringing their own voice to the page, don't need to be worried.
Yes, Caz. Has the agent simply asked to see some early drafts and their iterations? If someone saw mine, there would be no way AI could produce such a discombobulation. *Note - auto fill was used in the writing of the word 'discombobulation'. 🤷🏻♀️
A poor draft from a few years ago, then a great draft, nothing in between. That's odd. But why not request that evidence before offering the manuscript to the market? As you say, ask for the iterations.
I still find it difficult to believe that a LLM turned a weak manuscript into something amazing.
You highlight the need for nuances. I can see for critical/technical/legal articles or writing, but to have someone who uses anything and I mean anything to justify writing for self expression, creative, fictional etc is a little too much.
My feelings exactly. Submission to surveillance culture under the guise of I'm trying to get ahead of defending myself. To who? This is getting out of hand because of groupthink and hive mind mentality. And this is very anti-intellectualism and a pathway to authoritarianism in this day and age. There has to be some type of resistance. This isntreading on people's first amendment rights.
My exact feelings... As a non native English I combat this shit every time when I put through my text in grammarly, and then paste back to Substack, and it says 100% AI.. ffs
So, apparently Faladay does have evidence of his authorship and research. Let's see what happens.
Based on one of the interested editor's vague suspicions (12 publishers were interested in this book), the agents - or someone - ran his manuscript through Pangram, without his permission Yep, good old Pangram. The same detection tool used to kill Shy Girl publication.
Who benefits from this nonsense?
Did Pangram share prices go up again?
And again, the agents: he has the audit trail, they STILL failed to ask the right basic questions??!! WTF.
Yes, we'll see what happens. Definitely crazy. I hope he takes action. I hope he wins. I want to read his book!
It had a serious whiff to it from the get-go. Then the drip feed of what occurred. The agent's statement threw him under the bus. Then they implied he wouldn't be taking legal steps. Throwing him under a second bus. All a bit tawdry, designed to save their own skin. Agents with no moral compass.
Apparently he used AI to research matters relating to chemistry. Fine by me. As a Nigerian man, he also used AI to Americanise ordinary language, such as asking what Americans call a wardrobe. Again, fine by me. The latter is something a human editor would make him do. He anticipated this. Or maybe the agents gave that advice.
And I can't get past Pangram being the tool used on his manuscript, plus without his permission.
Again, for me, this points to the further hypocrisy here, because if he'd employed several humans to do all this work for him, there'd be no hoo-ha. For me, AI can be a great leveller in helping independent authors achieve what people with a lot of money are able to do, and I think that is what a lot of this is about.
How many decades have we all been relying on Microsoft to tell us where to put a coma, how to correctly use a semicolon, or, gasp, an em-dash? Thirty years?
If Faladay had purchased half a dozen chemistry books instead of using AI, would his manuscript have been materially different?
I heard from a little birdie today that they’re very well connected to journalists. Create a scandal, generate the hype, and suddenly they’re being written about everywhere. I suppose ruining people’s lives is good for business.
Actually, I hope he indie publishes his book and makes a fortune. And the idiots who made assumptions that he did something illegal or nefarious can go eat *hit.
Neela, it is a strange trade, having to log your own thinking to be believed. We keep something similar for build work: one line per session, what broke, what we kept, what we threw out.
It started as defense too, proof a decision wasn't just guessed at. But months in, it's usually the log we go back to first, not the code, because it's the only place that says why a thing is the way it is.
I work in tech, so I love the idea of the log as institutional memory. I think what bothers me the most is the reasoning behind all of this.
Huge difference between documenting your thinking because future-you will thank you versus documenting your thinking because someone may accuse you of cheating.
The first is helpful. Saved me many times. The second feels dystopian as hell. 😂
This is CRAZY! I’m actually part of the Anthropic settlement; they used four of my books to train Claude!!! I just ran my latest (100% human-written) post through Pangram and it said it was “91% human” and flagged one random paragraph… perhaps one that, oh I don’t know, SOUNDS LIKE SOMETHING I WROTE? I do keep meticulous backup but still, this is infuriating!!!
Thank you, Jenna, and I agree it is completely infuriating and basically doesn't work at all. I think one of the things I found is that if I include any form of citations or sources, it gets flagged as 100% AI, which is obviously insane because that seems to be indicating that we shouldn't be backing up our writing with references. Now, I know some people in BigTech would like that, but call me old-fashioned. I think any statistic needs to be backed up.
As a software engineer I can tell you first hand how flawed these AI detection algorithms are. I actually wrote an in depth article about it. I built my own simple algorithms and models to further prove the point that you should not blindly trust a detection score. I am open about how I use AI in my writing process. I use it to help me edit just like you do with a human editor.
People think you just punch in a prompt and get a polished novel or article when using AI. That is not true. AI is awful at long-form storytelling. It can’t keep the plot straight or remember what characters are doing or supposed to be doing. I have experimented with this as well just to see if it is even possible. It is definitely not possible.
I also keep all of my drafts from my book so I can show people if I had to. I honestly think we need to use our own judgment and stop letting AI tell us what was is writing by AI. It’s comical at this point.
I always enjoy reading your posts — thank you! I’m a college writing instructor, and I’m exhausted by provenance suspicion. Regarding version history: while there are extensions that mimic human typing, I’ve seen students do it the old-fashioned way (ha!) by using two devices. In class, I’ve watched students use an LLM on their phone and type the output into a Google Doc on a laptop. I’ve watched students use an LLM and painstakingly write the output by hand into a process journal I asked them to use in class — a process journal that could be presented as future evidence. 🙃
Writing is hard. Trying to make hard, uncomfortable things less hard feels profoundly human to me, so I don’t blame these students. I also believe the more they know about how these tools work, the more empowered they are. We learn about them, talk about them, and, yes, even use them as part of the class. The above examples occurred during a non-graded freewriting exercise for which I asked them to please not use any AI tools. 🙃🙃
I’m a visual artist and it’s a bit different for us. Some people are advising artists to make all their work on video to prove that it is human made. Videos can be generated by AI too, so I don’t think that’s a long term solution. Many artists have gone back to using physical media instead of digital art. For those using digital art, we are advised to save all of our layers. There is also the issue of AI reference images. AI can generate an image that pulls from another artist’s painting and then the second artist can use the AI reference image and unknowingly plagiarize the original artist. What I’ve been doing is going back to physical media primarily and using only my own reference photographs or ones published online before AI was around. Not to mention that the second my work is put online, it is scraped by AI. It is exhausting and it makes me just want to create artwork in isolation and never put it online.
This is a great article. Well-reasoned, well-written.
As to proving the human provenance of my own work at Forge City Noir, I have taken to "keeping my receipts" in a similar way that you prescribe. When I start a new story, I write out by hand a messy synopsis and/or outline. I put the working title and page number at the top, my name and the date on the bottom of each page.
When I like the outline and it has everything I want to talk about, I start writing the first draft by hand. As with the synopsis/outline, I put the title, page number, my name, and the date on each page.
When I'm ready to transcribe the story, I use a template that has a header and footer to contain all the provenance information, along with the version number. I print that then start marking it up. When I am ready to transcribe the edits, I save off a copy of v1 as v2, change the version in the header, then start applying my changes. Lather. Rinse. Repeat.
I use AI in my process to check historical facts and also as a writing coach. The prompt I developed gives the coach strict instructions to never, under any circumstances, offer me a rewrite. All it's allowed to do is point out an inconsistency, where the story is weak, a plot hole, etc. I'm human and I can't afford an editor, so I use the coach to catch things I miss, but all the work, from conception to final edit, is all mine. And I'm keeping the receipts to prove it.
My MC is a young woman. And when I have scenes in which she dresses 👗 up for an event, I describe what she's wearing (because it's first-person POV and she's a little fashion obsessed), which means I have to look up tons of outfits and accessories. Do I need to cite every source? If I Googled Paris Fashion Week, where in my novel do I list the link? Should I include footnotes or endnotes? How long does the reference section of my novel need to be?
This is a very fair comment. I think a lot of the time, I always think about academic writing, whereas you're absolutely right. In reality, every single source for every single idea, of course, does not need to be listed. Also, as the common thread through most of the comments in this piece, I really wish that this kind of provenance wasn't necessary at all and that instead authors were taking it at their word.
The thing that bothers me most about the AI conversation is that it feels like a revenue-driven issue. Really who cares how you made your story? It only matters when money is involved. Corporations don't like it because they feel cheated and worry about their bottom line. Indie authors don't like it for the same reason. Everyone is focused on publishing, not writing. Writers on social media won't talk about writing; they talk about their latest book. The whole issue is about who profits, not who wrote what. Who wrote what is a red herring dressed up like the argument.
Very well said, Gustavo. Sadly, I think it's because people in power are not happy about the fact that AI can actually be a great leveller. In many of these cases, it's going to be really interesting to see how this unfolds over the next 12 to 18 months.
It kills me that the onus is put on writers to prove they wrote their own work just because of detection tools that don't work very well, and are getting worse as the models get better at returning the "correct" next word.
This is ridiculous. Writers should not have to create a forensic record of their creative process in order to prove they wrote their own work. Drafts are messy. Some of us revise inside the same document. We write on different devices, in notebooks, in fragments, in our heads. We overwrite things. We lose things. That has always been part of writing.
Preserving whatever record naturally exists is sensible. But requiring writers to document their process in anticipation of an unreliable AI detector accusing them turns the presumption completely upside down. The burden should be on anyone making the accusation—not on writers to maintain a chain of custody for their own imagination.
And frankly, my creative process is private. I shouldn’t have to surrender it to prove that my work belongs to me.
I absolutely respect your right for this, Kelly, and honestly I wish this were the case. But sadly, many people are asking for provenance now in a way that just wasn't true even three or four years ago.
And that’s exactly what worries me. I understand that people are asking for provenance now. I’m questioning whether writers should normalize those demands by restructuring and documenting our creative processes around them. There’s a difference between preserving evidence that naturally exists and accepting that we owe anyone a running account of how the work came into being.
All developmental editors, copy editors, line editors, ghostwriters, and writing teams need to be named and publicly acknowledged as co-authors. It’s only fair that everyone know who actually wrote the book,
This is the argument I can make it over and over again. Thank you for also putting it into words. AI is just a great leveller, and I would like now all CEOs and alleged creators of work to tell me about their entire process: exactly what they wrote and what was created by ghostwriting instead. Only then can we start talking about AI detection.
I use AI, not in Substack but in other areas. It is an invaluable tool. A TOOL. For one thing I have stout ADD. AI removes a lot of the friction I have to deal with cognitively, such as organizing outbursts of thoughts and finding coherence in scattered observations. Neurotypicals cannot imagine the relief at finally being able to articulate what is swimming around in my head trying to get out.
And I use AI to reclaim my ability to create the artwork for my books. Arthritis has made the amount of work I create quite low. Ai toolset me do things like focus on the figure and let Midjourney or Photoshop fill in dozens of branches or blades of grass. Without that I’d be not creating.
Most of the time they are. That's who authors thank in the acknowledgement pages. But not as co-authors, because (unless they're gravely overstepping their bounds), that's not how editors work. (Ghostwriting and writing teams are a different story.)
Why don’t we start requiring authors to highlight which lines of dialogue are theirs and which ones were modified by a human editor? We don’t do that because it’s fucking absurd.
👏👏👏👏
people out there making millions with a ghostwriter
so how is hiring a ghostwriter any different than using an AI to produce a book? in terms of ethics. they do not always disclose if a ghostwriter was used. the only way you know is if they produce so many books in a short period of time, or if there is a "with" credit on the cover or "as told to" type of thing.
That’s literally what James Patterson does
And that's not illegal either.
So true and where is the outrage does Pangram have a detector for that?
Oh, there's outrage. It's just in the author communities, because readers don't understand how it works.
ghost writing has been happening for at least 100 years with books. HP Lovecraft ghostwrote a story. The Hardy Boys and Nancy Drew series were all ghostwritten. There are modern book series being ghostwritten which are obvious when they seem to be able to release a new novel every month and they've been doing it since before AI was a thing. I doubt Blake Pierce is writing all their own novels, for example
I’m aware of the history of ghostwriting.
Exactly. If one wanted to be outraged about anything, it should be that. Already-rich celebrities hitting the best-seller list for something they didn't write a word of.
At the end of the day AI a tool originally designed to help is being abused by the rich and powerful to keep the 1% on top and marginalize everyone else: they will always try to silence our voices, spreading the word is how we help each other speak out
But it's not working that way. Ordinary people ARE using AI for all kinds of helpful things. There is nothing wrong with that. AND, it is not illegal, nor is it anyone else's business. If they want to do that, let them. If you don't want to do that, then don't.
Truthfully, yes, AI is very helpful for a lot of people and however comfortable you are with using AI or not at all is entirely up to the individual. Sadly, like many societal issues where privacy should be paramount, “everyone” has an opinion and has to make sure “everyone else” knows it, even if they have no factual support or research, just rumor and conjecture. However, when patterns start to emerge like AI claiming that several successful writers from the same marginalized group where English is not their native language and flagging them to the point is costing successful and talented writers publishing deals, that’s not a simple fix, it’s a recurring problem that needs to be addressed before it becomes an even worse problem that xoukd lead to some really bad conflicts
I get really tired of the constant AI police.
Me too, Christopher, but we still need to fight them.
How would 'fighting' witchfinders realistically happen, given the discourse is about giving them what they want?
tell them to go pound sand, and self publish
I myself will no longer be reading self published works published after 2020 (thought it might go to 2018) because authors are simply not honest about their use of ai, and because the use of ai goes far beyond simply whether its your voice, far beyond the fact that its theft (in fact I'll probably be blocking this blog because it uses ai images, but I respect the writer for disclosing so I can make that choice), and far beyond the environmental impacts.
I wish those that champion it would do a much deeper dive on the ways it is harming so many people and the future and stability of civilization.
I'm sure others are not far behind in ditching books written during a certain time period because the trust is simply not there. So feel free to tell the ai police to pound sand and self publish, but I recommend when you do, you at least be as honest with your readers, as the author implores.
In the context of raging AI debates, I offer this Substack PSA:
<SUPPORT YOUR LOCAL POET> I self-published two books of poetry--poetry I wrote, read aloud at open mics, edited, rewrote, shared with other writers, revised, performed, collected, and proofread. A book designer created and uploaded the cover and contents files. Printed on demand, the books are beautiful and still selling, one here, one there. If you would like to purchase a non-AI collection of smart, accessible, well-written, meaningful poetry--not the kind you have to decipher, but the kind you read, think about, and read again--DM me for a signed copy. Or find *WHEN EAST WAS NORTH* and *RABBIT SUN, LOTUS MOON* on Amazon, by me, Andrea Millenson Penner. Or find my 2024 chapbook, *Upcycled: Poetry Repurposed* new and updated work because "like people, poems deserve a second chance." Published as Andi Penner. Thank you for supporting human-generated creativity. Sent with love and gratitude.
I don't care if someone has used AI or not. I doubt I've ever read a GenAI novel as I nearly exclusively read works from the 20th century and earlier. But if I'm reading an article and suspect, or later find out, that it's AI - so be it. Was it informative? Was it engaging? Was it clear? Did it seem accurate and truthful? That's all that matters to me.
This right here is why I hate GenAI. I began working on my debut novel in 2015. I paused to get an art degree so I could retain as much creative control over my novels as possible by doing all my own covers and marketing for self-publishing. I do not use AI. I roll back updates and disable every AI "tool" every time I can. But it doesn't matter that I have all the receipts to prove I wrote my novels word-by-word. I can't go back in time to publish it before 2020, so it's assumed I used AI and am lying about it. I hate this timeline!
Social media has been doing (in many regards) enormous harm to people and civilisation for nearly 20 years now but that hasn’t stopped millions from using it all day long.
Wow! Thank you for putting so succinctly what I have been thinking. Adding to the issues you mentioned, it's the tone of the article or a piece that is annoyingly similar to other AI generated piece. It feels like waking up to someone's shrill voice and hearing that all through the day, over and over again!
Gordon Lish and his overbearing editorial control of Raymond Carver's entire body of work is now considered problematic, but that took decades. His books are still available.
I’m not familiar with his work.
He's quite famous for his short story collections. Lish's career is equally infamous.
"Lish's cuts were sweeping, wholesale, and so violent that Carver described them as "surgical amputation and transplant". He responded with a now-famous cri de coeur, writing a letter to Lish explaining his horror: "I don't want to sound melodramatic," he wrote, "but I've come back from the grave here to start writing stories once more . . . I'll tell you the truth, my very sanity is on the line.""
https://www.theguardian.com/books/2009/oct/24/ernest-hemingway-raymond-carver?CMP=Share_AndroidApp_Other
I absolutely adore Raymond Carver. I spent a long period of time trying to emulate his writing with my own short stories. Nothing as obvious as plagiarism, but someone that I greatly admired.
Yes, I love his writing, or what was left of it. A unique voice, difficult to emulate. Writing style and subject choices for fiction are much like finger prints.
I was gutted when the Lish controversy became public, more gutted for Carver than myself as a reader. I wondered why Lish didn't stick to his own writing and spare other authors his editing.
Imagine having an editor who changes the entire intent of your writing.
It is the very reason I gave up on human editors. This is impossible to explain to people that have already decided that there is only one way to leverage AI when writing, and that is copy and paste, which is absurd.
In the same way that one can ignore a human editors suggestion you are free to ignore the AIs editorial suggestions as well, it is really that simple. That is how you keep your voice.
Exactly!
Might as well record videos of your entire workflow while you're at it
There seems to be a growing crowd whose knee-jerk response to anything decently written is "AI slop" or "Thanks, ChatGPT"
Trying to appease such people is a waste of time for everyone, frankly...
100%
Right, and the post above actually gives a way to do that without turning it into a disclosure fight: a per-session log of what you asked for, what you took, and what you rejected. Nobody publishes that log, it just sits with you. And the rejected column ends up being the interesting one over time, it is not just a record of edits you skipped, it is a pattern of exactly where your judgment and the model's default keep splitting, which tells you more about your own voice than any single accepted line does.
Thank you. I wonder why AI is so bad? I use it frequently in my writing. I feed it information, it gives me sentences and paragraphs. I accept or reject and request a rewrite. I feel like it is only a tool to help my work better and faster. If a book is good, I hardly think AI wrote it from a single prompt!
Speaking of, how do we know, as authors, that the human editors don’t use AI for edits, inadvertently sabotaging the work without our knowledge?
They frequently do, which is how authors get caught out. The same with commissioned custom art, particularly when it's cheap. That said, I don't see how a GenAI artwork cover is any worse than using the same stock photo as thousands of other authors have, which happens all the time in genres like Romance. I've bought stock photos for covers before and I've seen at least one of them used by another author. You can buy exclusive licences from some sites, where they then remove them from further sale, but why should we? As a reader I couldn't care less about it.
I completely agree. I think the issue is licensed resale/reuse versus nonsanctioned edits of an original work without the authors acknowledgement or consent AND/or AI used under the auspices of original work. It changes the landscape of creative and original work. I’m not against AI, either. I can just understand the argument on both sides.
Well, then they would have to maintain their massive audit trail as well
Thank you for this.
I am so torn about this topic. On the one hand, I‘m a literary translator and will likely lose my job because publishers and authors are unwilling to pay me a living wage anymore, and I detest dishonesty and plagiarism. On the other hand, I‘ve read many amazing stories in the realm of amateur fiction that would never be picked up by a publishing house and for successful self-publishing would have required the work of an editor who these gifted writers could never afford. So like the options for self-publishing I see this AI as an important step towards making publishing more equally accessible. I‘ve also encountered a creator who had AI write his complete novel. He was so generous to share his complete, painstaking process from building the plot and the characters to perfecting the finest nuances of voice and style. No, he didn‘t write the novel, but he… was the architect responsible for the placement of every word. The product was a fascinating read that shouldn‘t be dismissed as a literary creation. I‘m just sad for the creator that he missed out on the experience of writing and authorship.
In terms of getting published, I must admit that I am deeply discouraged. Even before the advent of AI the way authors increasingly had to put in massive efforts in terms of social media and how "shitstorms" had basically become a given and pretty much unavoidable made the idea of publishing less attractive. And now this…
What can be a reasonable way forward?
Your experience provides lots of food for thought. It's definitely not a completely black or white situation; there's plenty of room for multiple grey shades in the middle. At the end a tool is just a tool and it's in the way we use it, like fire, like a knife, like the wheel. Steph, I too am not sure what the way forward is, but what scares me is not artificial intelligence. It's human idiocy, and our proven tendency throughout our many civilizations to self-destruct, instead of learning to live harmoniously and cooperatively.
I definitely agree with you there human stupidity is pretty terrifying. What’s even more terrifying to me is that I fear that this is just how we are. This is what we are as beings and that we won’t change. Maybe we can’t change and if that’s the case that’s sad.
The way forward is to choose to break free of the old world that does not exist anymore and embrace the new world.
Trad-publishing is over, and it is time to move on.
https://kennetheharrell.substack.com/p/the-age-of-publishing-gatekeepers
I have the outline, only, of a book I wish to write, based upon personal experiences. I do not intend to use any AI, or grammarly at all. I do intend to use a human editor. I have zero interest in these tools. Thank you for this information. I will be saving my paper trail.
will you be disclosing anything your editor writes that isn't your actual writing?
Of course, no. I was edited once, by my own mother. In junior high, she typed up a short story I had written for class. She changed a sentence I had written from “Wait for me!” to “Wait! Oh wait for me!” It was too late to change it back. My typing would have taken another two days, with eight false starts. I tell you this to say her editing one word without me agreeing to it is still under my skin after 50 years. So it is unlikely anyone will be significantly editing anything other than basic grammar. I will take suggestions, but it will require discipline to avoid acting on that same childish feeling when it wells up. Sad when emotional responses fail to evolve.
Do you do that on the human editors who you pay thousands to to change it up? Yet that’s considered “professional.”
Nah because you can - because that's called a saved copy and tracked changes, comments turned on.
Ready to prove my proof of process… we have been ready since we learnt ai has stolen our text… with enough to take multimillion dollar companies to court.
Also, that means a court date, coz I shall be compensated for my time :)
This will calm down in the end because have they noticed? Publishing is based on automated processes
I'd be interested in a statement from the publisher as to their objection to minor use of AI. If their stance is that it's not the author''s own work, I don't see how that is defensible. Using Grammarly means it's not all the author's own work. Ditto using a copy editor. Ditto asking friends to read a draft and give feedback. And using a ghost writer?
Maybe it's time for Arthur Miller to rewrite The Crucible, replacing witchcraft with AI.
This is exactly how I feel as well, Martin. Similarly, for the extent to which tools are used for editing, even in Microsoft Word, we've genuinely got to a point of insanity at this point, and you're absolutely right to point to Miller's work is a very depressing touch point.
I can see typewriters coming back. Think of all those books pre digital written by hand and edited by hand. Authors have become lazy and AI is the ultimate laziness. Challenge is where AI is used to edit (that is tweaking) versus it being used to write chapters that are then edited by humans to make it less AI written. Maybe we will see disclaimers in books ... this book was drafted on paper with an ink pen and here is a video of them at 3 in the morning writing chapter 6 etc. I want my authors to have struggled in their writing ... thats what makes them authentic but it seems from the research most readers dont care as long as its a good read
Let's bring back the abacus too, because obviously mathematicians are getting a little lazy, letting AI solve proofs.
Who gives a fuck. The only criteria is whether or not its a good read
Do you think authors who use ai don’t struggle? You are mistaken. Why is anyone writing their replies here on social media on the internet? I propose henceforth we go offline and correspond only by handwritten notes delivered on horseback (only by men because clearly women don’t belong on horseback).
Yes, the bon mot I learned was: Good reading comes from hard writing.
Martin's right that nobody tracks Grammarly or a good copyeditor the same way. The distinction the post's own fix points at isn't purity, it's disclosure. An audit trail log doesn't ask whether a tool touched the work, it asks what you asked for and what you kept versus threw out.
That's a different question than "did AI touch this," and it's the one a publisher can actually check, since the log just shows who made the final calls on the page.
But then a copy editor is another professional who is doing a job and getting paid for it and contributing the publishing industry. Using AI removes that cog and destabilises an already fragile and vulnerable system in my opinion (but I’m not an academic, and have never published a book - just an enthusiastic reader who’s watched my own industry crash and burn thanks to AI)
And writers likely pay to use AI. Virtually impossible to write a book and keep track of everything in free versions. And any AI would lose track of characters, story arc, etc. without human intervention (at least for now). It struggles to write a cohesive chapter without guidance.
How editors work is they suggest changes. They do not make changes - i think even a semicolon instead of a period would have a comment as to why, and that is with tracked changes on even if they did just go ahead and fix a comma placement. (One would hope to do their own line editing before it gets to a human editor who charges by the word. Imagine an editor having to fix your commas routinely.)
Anything that inserts words or changed them around is not being an editor, it is doing the writing…
Which is required to be disclosed.
A dumb spelling and grammar checker like in windows 95, is a tool suggesting the change of affect to effect. Links, the grammar assistant cat in windows 95 has a pop up bubble that says “did you mean effect?” And gives a little explanation as to why and the definition.
I haven't a clue about grammarly, but nothing and no one should be changing words around. It, or the human editor, should only offer suggestions for the author to decide to implement or not. I think a lot of people do not know what editors do and what they don't do.
That's not quite right about editing. When I edited journals, I changed sentences around if they were eg. from European authors who had used European sentence structure. In fact I edited everything that needed editing to make it sound like UK English.
The onky thing I checked with authors was things in equations such as if an *x* needed to be italicised.
Interesting! I never had an editor move things around for me or change my sentence structures at all. Comments only, then comma corrections. When I did a cert 4 in professional writing and editing with Tafe gippsland and then diploma with university of new england... We were taught to be very non intrusive, as it wasn't our work. 🤔 Interesting.
Correct. I use grammarly to edit written legal arguments. It doesn’t just check spelling it suggests words, phrases and punctuation.
Or just using any word processor
Using a spell checker is A I.
Yesterday, i cut and pasted the first page of a celebrated 1974 book called The Dispossessed by noted author Ursula Le Guin into an AI and asked if it was AI assisted. I did this because I was shocked by the first two lines and each subsequent line because that page contained ALL of the major AI tells. Line after line after line. The LLMs borrowed heavily from her prose. The AI came back with 75-85% AI and a few pages of all of the justification given in various articles about contest winning and publication receiving authors outed as using AI. But of course this work is not a product of AI. 🤖
So the real objection to AI seems to be that AI is copying the style of established (or any and all writers) and offering that style to anyone using the AI. True?
Correct. The first page of that book is the most perfect 1:1 representation of what LLMs copied from an author to structure its output. The tail is now wagging the dog.
I will never pander to bad-faith accusations of AI use, and I will never keep an audit trail. The work *is* the audit. How it got onto the page is no one's business. Whether a suspicion machine like Pangram likes it or not is irrelevant. AI is a research/writing tool and a universal translator for human ideas. It was trained on the best of human writing. But raw model output is generally unsuitable for use without extensive human editing anyway. Which is how it should be. Every author must stand behind their work regardless of their process.
Eventually this controversy will die a well-deserved death and people can get back to judging writing on craft, along with quality of thought and argument.
This is my sincere hope as well, Sean.
I feel the exact same way. This nonsense is all the fault of the anti-AI crowd who care less about content than they do about sticking their noses in someone else's business. It shouldn't matter how the words got on the page if you enjoy reading them. I've never once asked myself about a favorite author, hmm i wonder if all this writing is theirs or has it been extensively rewritten by editors. Because I don't care. I only care that I enjoy reading the book.
👏👏👏
I totally agree. How is this different than that of an editor, copywriter, or ghost writer? When I buy a book, I'm not doing an over-arching detailed analysis of the words on the page. Honestly, I don't care. I will pick up the occasional typo because secretly I'm a grammar-nazi, but whether a turn of phrase is created by the author, editor, or AI - doesn't concern me.
Exactly THAT 👏🏻👏🏻👏🏻
I'm anti-mimmix (AI), but don't report AI slop. It's a useless endeavor, as that's part of training mimmix & removing human authorship from the household. I refuse to be part of either end of that.
You don't? You don't think about the moral lessons in a story? Or the real life stories that ground what has been written on the page? You don't wonder what an author meant by something? Or if it was changed or modified by an editor? You don't find joy in seeking out what they as a person was attempting to say? You think only of its purely entertainment value? You don't wonder about your favorite authors as people and how that impacts what you're reading?
No. I don't write or think that way so why should i read a bunch of nonsense into other people's writing?
You don't write with intentions to convey an idea? Or based off of lived experience? That's fascinating and quiet unique
Lol it's really not. People who don't write fiction think it is because they believe their high school Lit teacher who designed entire classes on figuring out what an author meant by something. It's actually pretty funny. All the patterns in my fiction put themselves there without my conscious thought. But lol non-writers think we're constantly threading in our own values and lives into our fiction and making esoteric points through metaphors or whatever. Maybe some do, but I'm not one of them, and i doubt I'm alone. The things I write about are very far removed from my actual life or experiences. That's why it's called FICTION and specifically, speculative fiction. Imagination means you can imagine lives that aren't your own.
This is actually a fascinating claim. You believe all your writing has been written without conscious thought. That lessons, habits and biases of your own life have been left out. While you sit down to write who do you envision is writing on your behalf? Is this same source writing for all others? Is it only in fiction where this occurs?
The reason I keep an automated audit trail is so I can roll back misguided changes and keep track of why I made certain decisions. That just happens to be an excellent audit trail in general.
This is my sincere hope! My words, my thoughts, my voice. Who the hell cares how it got on the page?
Right? It’s like people think book writing is about generating text and not something deeper.
The AI police are constantly talking about how bad AI writing is but they rarely define what “ai written” means. Is it okay to use it to edit? Is the problem generating original content with the tool? They don’t know. All they say is that AI writing is slop. But if it is so bad, it should stand out like a sore thumb. And it frankly doesn’t in most cases. Because in most cases authors who are using it are using it as a tool. They are taking some of what it gives, rejecting other stuff, and heavily editing anything it produces. I completely agree, let the prose stand for itself. It’s not my business or anyone else’s how it got there.
Publishers are doing this to avoid lawsuits. They don’t want to get accused of copyright infringement. I understand protecting themselves from a business standpoint, but they’d do better to define their terms up front and add an AI indemnity clause to the contract.
Anything creative requires tools. If the creator using the tool is mindful, and allows common sense and wisdom to guide their decisions in crafting a work — whether it is a brush on canvas, hands on a drum, or fingers on a keyboard — what interests me is the final piece. Goodness knows I spend just as much time babysitting Claude and correcting it than I do crafting what I write so the written word remains true to my voice. I care about the craft and interior process of thought, memory and inspiration deeply. I am also immensely grateful for the edits and audits available to me, a true gift as a pen-mute woman who disappeared from life for seven years and needed a way back to words. I never could have imagined what form that support would take.
I mean it matters because writing isn't judged primarily off of craft but off of what one person is attempting to say to another. When we add more and more things between those two people (like AI) we're very naturally going to run into issues…
I am not doing any of this BS. According to my stats no one really reads my work anyway. This seems like double the work for the same amount of non-readers and non-fans. My advice, just go read something else. I honestly do not care.
I absolutely respect your right to say this, Kenneth, and I genuinely believe that within 12 to 18 months we'll be entering a phase where a lot of this will be by the way anyway. At the end of the day, all AI really is is a selection of tools that, when harnessed correctly, can massively improve the quality of writing.
You are in good and plentiful company when you say AI "can massively improve the quality of writing."
I, however, must have a different definition of "quality." If the grammar is perfect but the writing is now bland, lacking any enjoyable voice or personality, I would not say that the quality is improved.
AI writing that's so obviously AI, no need for a detector, is mind-numbingly boring, and most importantly, it's really sad. I know writers whose content is interesting, and they say "I write it and then ask AI to rearrange and tighten it up." What they're actually doing is using AI to strip out any interesting parts of themselves from the writing, their own human quirks and thoughts, in order to make it "better."
That is a loss, for both writer and reader. That is why I wish writers wouldn't use AI. I will never reread their books the way I reread the authors whose voice is delightful, or fascinating, some books so many dozens of times that the binding falls apart. But that is how I define "quality" in writing. And it is something AI will never be able to do.
I respect your opinion, but I have to ask: to what extent is getting editorial assistance from AI any different from that of a human?
This is a great question, for me, the difference is that an AI editor does not decide what my motivations must have been before reading what I actually wrote. Using AI as an editor helps me focus more on narrative structure and refinement, not managing the personal biases and ideological imperatives of a human editor. I don’t have to justify what I choose to write about.
AI editors do not turn every editing session into a long back and forth lecture about what a “Male” or “Black male” writer “should think” and write about when it comes to issues of race, gender, or politics. I don’t need lectures about “the patriarchy” and “male privilege.” I don’t need lectures about what I do or don’t have the right to talk about because of who or what I am.
When I put hands to the keyboard THAT is what gives me the right, I have the right because I take it! I don’t need anyone’s permission to create a character or address a topic of my choosing. What I want from an editor is actual engagement with my work, and that is what I get from using AI as an editor. Example;
Where can pacing be improved?
Does this scene make sense?
Does this characters argument make sense?
Is the prose clear?
Is this order of actions in this scene clear?
What can be improved in this dialogue exchange?
Are the character’s actions or reactions believable given the way this scene is set up?
Where am I repeating myself?
Have I failed to explain something clearly?
In my view this is the advantage of using an AI as an editor over tolerating a human one.
Very depressing to read this.
Human editors and publishers should have treated writers better. The publishing industry brought this on themselves.
Why is it so easy to tell when an AI edited something? That's the essence of the difference between human editors and AI.
It is heavy-handed in specific ways-- it pulls all writing in line with a generic standard on specific lines- and by the way, pulling it up to standard, or down to standard- and makes it all sound vaguely the same.
It doesn't know any better. It can't. It's programmed with an as-yet-limited hierarchy of "these are best- good- acceptable ways of doing this" and that's the model to which it sets any writing.
Maybe this will become less obvious over time, but until AI has a soul, it won't be able to have an author's voice.
Human editors, in general, do not do this to such a broad extent. Not to claim structure is never addressed by humans, but they don't make everyone's writing sound the same, with the same structure and same generic catch-phrases. Because it's a human, looking at your work with their human brain and making it better, not indiscriminately molding it to a predetermined generic standard.
I loved this comment. And I agree. I actually had to be very stern with the AI I finally chose to help with my writing and make it clear that in no way shape or form is it to ever even suggest an edit that would follow modern writing standards of tailoring to the weakest literacy link or strip my writing to read like boring corporate internal memos. My audience is not children's lit and my literary role models that I try to emulate are King, Koontz, Twain, Hemingway, Browning, etc. Literacy of that caliber. I may not be as great as those writers, but they are the masters at storytelling. I refuse to let the AI try to strip that from my writing for the sake of making it easier to read.
I have even contemplated pulling my fantasy from the one website where I share it simply because my readership fell off the minute I took down my first draft and posted up my second draft. The readers seamed to prefer the stripped down, less immersive version I expanded it from. I often wonder what happened to all the readers like me who love being able to feel fully immersed and lost in the story.
That is all well and find but again, there is not just one way to use this technology. People that assume that are just misinformed. Some of my earliest posts explain multiple ways writers can use this technology. (see below) I think I have covered precisely how I use AI pretty well, so I really feel no obligation to explain it even more.
The reality is that I gave up on human editors years ago. Once you sit through enough finger-wagging lectures by ideologically driven editors, you reach a point where enough is enough. I have found that human editors assume your background and writing motivations based on their biases. This makes the ease and practicality of using AI for feedback, editing, proofreading, and story structure checks much more appealing.
If anyone has a problem with that, or my work the block feature is at their disposal.
My Writing Process: https://kennetheharrell.substack.com/p/my-writing-process
Using AI Art as Gumbo Starter for the Imagination: https://kennetheharrell.substack.com/p/using-ai-art-as-gumbo-starter-for
Using AI as a Reality Simulator: https://kennetheharrell.substack.com/p/using-ai-as-a-reality-simulator
Part 2 Tweaking the AI Reality Simulator. An Ongoing Experiment in AI Simulation: https://kennetheharrell.substack.com/p/part-2-tweaking-the-ai-reality-simulator
Part 3 Interview With An “AI Entity” in Simulation using a Custom GPT: https://kennetheharrell.substack.com/p/part-3-interview-with-an-ai-entity
Using AI To Explore Ideas For Science Fiction Stories: https://kennetheharrell.substack.com/p/using-ai-to-explore-ideas-for-science
Part 1. I Had My Custom GPT Do An Analysis of All My Work. I Found its Insights Quite Fascinating: https://kennetheharrell.substack.com/p/part-1-i-had-my-custom-gpt-do-an
The age of publishing gatekeepers is dead: https://kennetheharrell.substack.com/p/the-age-of-publishing-gatekeepers
I write something, ask AI for critique, then discard half to two thirds of its suggestions because, yes, they would sand off the edges and make it no longer sound like me. But the third to a half that I consider and then implement my own way do I prove the work. It's a tool. And like any tool, requires discernment in use.
I don't see what's wrong with it. If the article written by AI gives the information that the writer has asked for and suggested what does it matter?
People that have nothing better to do worry about it. Then again, I’m sure they look at every photo with tears in their eyes wondering whether or not Photoshop was used on that photo before they get to enjoy it.
Nope. It is manipulative, overwhelmingly wrong, stupid & invasive. Ban it.
I’m 100% with you on this. I’m not jumping through hoops to convince someone I don’t know (if they’re even a human being) my work is genuine.
I’m just going to sit back and wait until the bubble bursts, and not waste my energy on this witch hunt.
It really is a witch hunt which is precisely why we need to do what people during the witch trials should’ve done and that’s to tell these people to #uck off!
The problem isn't human editors. The problem is the BS editing by commitee to not offend anyone. F them. Write your book.
Right. We shouldn’t have to do double the work just so a robot MIGHT label our work correctly
The old AI tells are becoming pretty useless. Once everyone knows the words, rhythms and habits to avoid, the models learn them too.
This is why I’ve stopped caring quite so much about detecting AI from the prose. That test has a very short shelf life.
My name on something means I’m responsible for what it says, however I made it.
The harder test is judgement: is there an actual idea here, is it mine, and is it anything more than reheated internet tripe?
Absolutely this Daniel. And this is what will always differentiate excellent writing.
Exactly. The ideas are mine however it’s written.
The sad part is a lot of people are just going to be lousy on purpose because they’re afraid that if they write the way that they normally do (with many of those telltale signs of AI which quite frankly are just part of human writing) they’ll get flagged and shamed for using AI.
Why don’t we take a lesson from the witch trials and simply tell these people to fuck off instead of making complicated statements about how we did NOT fornicate with the Devil?
To hell with the how I made this statements. To hell with the auditing trail for people who are strangely more concerned about tool usage than the content of a persons work.
Well, if the current American president has taught us anything, it's that no one can shame you without your consent.
I am not particularly political, and I have already written about ways in which writers can use AI. Some of my earliest posts talked about this topic and I think I have covered precisely how I use AI pretty well, so I really feel no obligation to explain it even more to people that would otherwise not even read my work anyway.
Again, if anyone has a problem with that, the block feature is at their disposal.
My Writing Process: https://kennetheharrell.substack.com/p/my-writing-process
Using AI Art as Gumbo Starter for the Imagination: https://kennetheharrell.substack.com/p/using-ai-art-as-gumbo-starter-for
Using AI as a Reality Simulator: https://kennetheharrell.substack.com/p/using-ai-as-a-reality-simulator
Part 2 Tweaking the AI Reality Simulator. An Ongoing Experiment in AI Simulation: https://kennetheharrell.substack.com/p/part-2-tweaking-the-ai-reality-simulator
Part 3 Interview With An “AI Entity” in Simulation using a Custom GPT: https://kennetheharrell.substack.com/p/part-3-interview-with-an-ai-entity
Using AI To Explore Ideas For Science Fiction Stories: https://kennetheharrell.substack.com/p/using-ai-to-explore-ideas-for-science
Part 1. I Had My Custom GPT Do An Analysis of All My Work. I Found its Insights Quite Fascinating: https://kennetheharrell.substack.com/p/part-1-i-had-my-custom-gpt-do-an
The age of publishing gatekeepers is dead: https://kennetheharrell.substack.com/p/the-age-of-publishing-gatekeepers
Looks like I was caught being autistic in public again. I meant no offense. I actually agree with you, and I was just pointing out that if people want to shame me for using AI, good luck with that because I'm not ashamed. I don't buy into their frame about any AI use tainting the entire work. As long as I write all the words myself, which I'm going to do because that's the fun part, they can go to hell.
The audit trail idea is practical, but it’s also a little heartbreaking that writers now need a paper trail to prove they created something. It makes me want to stop writing and go back to journaling :(
Thank you Neela. And I absolutely agree with you. So sad that it has come to this. 😢
Does any serious writer write a poem let alone an entire novel without saving versions, research, notes, edits? There would always be an entire cache of material. I can't imagine anyone blithely overwriting every version of their drafts, or tossing away research once used. Even in business, people save versions.
There's so much wrong with this lastest episode, too much to go into. But at the top would be the lack of due diligence PRIOR to selling a manuscript, and if LLMs can churn out a novel so compelling that it's worth $2M up front, wouldn't we all be doing it? We're not, because LLMs can't produce any fiction that good.
I don't keep drafts of my writing when I rewrite. The clutter bothers me and it's like building on a existing foundation so I have the old memorized and only needing the new changes. Although most of my world building is in my head and done verbally when explaining my story to someone. I guess the main audit trail I would have is that I listen my different iterations so I have a lot of elevenreader files. Although for me I'll tell them here's my writing I want to see you FORCE AI to write like me because I know it can't.
Great if that works for you!
If I write a decent sentence or an awful sentence, I know I can't recreate either if I lost them, let alone entire chapters. 😂
Electronic versions can definitely feel like clutter, but using a standard version format helps.
Honestly for most your statement would be perfectly accurate, but I feel as if the desire to publish and having to confine your writing to a box to fit through the gate has made people forget that's it's Art. That's it's creative and free like water and from person to person it's all different and SHOULD BE that's what makes it amazing. If we all follow the same process and patterns then we really aren't much different from AI in my opinion.
😂 Overall I've learned my mind works oddly so while I agree with your answer I'm always worried about the outliners .
I am careful with document control at work as a public servant. But I confess I do a lot of overwriting for my Substack series. I've only recently started seriously considering turning my work into a book down the track. So I appreciate the advice.
Ah, yep, all too familiar with document management in the public service!
I definitely don't miss the torture of dueling edit tracking. 😂
I don't use edit tracking on my own work, I would find that weird, but I do create an electronic filing structure, and save versions under individual chapter files, or research topics, etc, using the normal v.xxx format and a date as part of my naming convention.
I agree with you, Caz. And if I ever finish the book I've been threatening to write since 2002, there will be plenty of evidence. 😂
But I was coming at this from the perspective of a nobody writer on Substack (me personally).
I'm thinking defensively. What happens when an ordinary writer gets accused of using AI?
That's the part I find depressing.
I also agree that the publishing world needs to check itself because I genuinely believe this author was wronged on so many levels. The problem is his name has already been tarnished.
Agree, and this writer was a first time writer (as was Shy Girl ?), so an ordinary writer, if you like, until it all went pear-shaped with a $2M deal. From the little I've seen, the agents withdrew the manuscript, it sounds like he had no audit trail to share with them (why didn't they ask before putting it up for auction?) He first drafted the book several years ago, apparently subpar. The next step was a great manuscript. Nothing in between.
I would hyperventilate if I didn't have every version. I know that's not how everyone works. Oddly, I think people really putting in the work, bringing their own voice to the page, don't need to be worried.
And even if they someday can, best of luck marketing it in a market already saturated before AI came along
Yes, Caz. Has the agent simply asked to see some early drafts and their iterations? If someone saw mine, there would be no way AI could produce such a discombobulation. *Note - auto fill was used in the writing of the word 'discombobulation'. 🤷🏻♀️
Yep, that was my thought.
A poor draft from a few years ago, then a great draft, nothing in between. That's odd. But why not request that evidence before offering the manuscript to the market? As you say, ask for the iterations.
I still find it difficult to believe that a LLM turned a weak manuscript into something amazing.
You highlight the need for nuances. I can see for critical/technical/legal articles or writing, but to have someone who uses anything and I mean anything to justify writing for self expression, creative, fictional etc is a little too much.
Next brain scans will be required.
My feelings exactly. Submission to surveillance culture under the guise of I'm trying to get ahead of defending myself. To who? This is getting out of hand because of groupthink and hive mind mentality. And this is very anti-intellectualism and a pathway to authoritarianism in this day and age. There has to be some type of resistance. This isntreading on people's first amendment rights.
Exactly!! I am not giving more and more and more of my life over to extreme surveillance
My exact feelings... As a non native English I combat this shit every time when I put through my text in grammarly, and then paste back to Substack, and it says 100% AI.. ffs
So, apparently Faladay does have evidence of his authorship and research. Let's see what happens.
Based on one of the interested editor's vague suspicions (12 publishers were interested in this book), the agents - or someone - ran his manuscript through Pangram, without his permission Yep, good old Pangram. The same detection tool used to kill Shy Girl publication.
Who benefits from this nonsense?
Did Pangram share prices go up again?
And again, the agents: he has the audit trail, they STILL failed to ask the right basic questions??!! WTF.
No way. This is crazy!! Looking forward to the litigation that follows and also the best selling book!
Yes, we'll see what happens. Definitely crazy. I hope he takes action. I hope he wins. I want to read his book!
It had a serious whiff to it from the get-go. Then the drip feed of what occurred. The agent's statement threw him under the bus. Then they implied he wouldn't be taking legal steps. Throwing him under a second bus. All a bit tawdry, designed to save their own skin. Agents with no moral compass.
Apparently he used AI to research matters relating to chemistry. Fine by me. As a Nigerian man, he also used AI to Americanise ordinary language, such as asking what Americans call a wardrobe. Again, fine by me. The latter is something a human editor would make him do. He anticipated this. Or maybe the agents gave that advice.
And I can't get past Pangram being the tool used on his manuscript, plus without his permission.
I think I hate Pangram. 😂
Again, for me, this points to the further hypocrisy here, because if he'd employed several humans to do all this work for him, there'd be no hoo-ha. For me, AI can be a great leveller in helping independent authors achieve what people with a lot of money are able to do, and I think that is what a lot of this is about.
Yep.
How many decades have we all been relying on Microsoft to tell us where to put a coma, how to correctly use a semicolon, or, gasp, an em-dash? Thirty years?
If Faladay had purchased half a dozen chemistry books instead of using AI, would his manuscript have been materially different?
Well, it's also about Pangrams market value. They must be loving this stuff.
I heard from a little birdie today that they’re very well connected to journalists. Create a scandal, generate the hype, and suddenly they’re being written about everywhere. I suppose ruining people’s lives is good for business.
Actually, I hope he indie publishes his book and makes a fortune. And the idiots who made assumptions that he did something illegal or nefarious can go eat *hit.
I look forward to that, Kieran.
I cannot wait to buy his book :)
Agree, all the better if self published.
Unfortunately he would still need to prove that it wasn't written by AI, because AI generated text can't be copyrighted, it's not owned.
And no, I don't understand how Amazon is supposedly overrun with AI created books when their own system is supposed to block AI slop.
Neela, it is a strange trade, having to log your own thinking to be believed. We keep something similar for build work: one line per session, what broke, what we kept, what we threw out.
It started as defense too, proof a decision wasn't just guessed at. But months in, it's usually the log we go back to first, not the code, because it's the only place that says why a thing is the way it is.
I work in tech, so I love the idea of the log as institutional memory. I think what bothers me the most is the reasoning behind all of this.
Huge difference between documenting your thinking because future-you will thank you versus documenting your thinking because someone may accuse you of cheating.
The first is helpful. Saved me many times. The second feels dystopian as hell. 😂
This is CRAZY! I’m actually part of the Anthropic settlement; they used four of my books to train Claude!!! I just ran my latest (100% human-written) post through Pangram and it said it was “91% human” and flagged one random paragraph… perhaps one that, oh I don’t know, SOUNDS LIKE SOMETHING I WROTE? I do keep meticulous backup but still, this is infuriating!!!
Thank you, Jenna, and I agree it is completely infuriating and basically doesn't work at all. I think one of the things I found is that if I include any form of citations or sources, it gets flagged as 100% AI, which is obviously insane because that seems to be indicating that we shouldn't be backing up our writing with references. Now, I know some people in BigTech would like that, but call me old-fashioned. I think any statistic needs to be backed up.
I'm running out of words for "absurd." SMH. Thanks for this detailed dive!
As a software engineer I can tell you first hand how flawed these AI detection algorithms are. I actually wrote an in depth article about it. I built my own simple algorithms and models to further prove the point that you should not blindly trust a detection score. I am open about how I use AI in my writing process. I use it to help me edit just like you do with a human editor.
People think you just punch in a prompt and get a polished novel or article when using AI. That is not true. AI is awful at long-form storytelling. It can’t keep the plot straight or remember what characters are doing or supposed to be doing. I have experimented with this as well just to see if it is even possible. It is definitely not possible.
I also keep all of my drafts from my book so I can show people if I had to. I honestly think we need to use our own judgment and stop letting AI tell us what was is writing by AI. It’s comical at this point.
Thanks, Joshua, for speaking sense. Comical is exactly where we are right now.
I always enjoy reading your posts — thank you! I’m a college writing instructor, and I’m exhausted by provenance suspicion. Regarding version history: while there are extensions that mimic human typing, I’ve seen students do it the old-fashioned way (ha!) by using two devices. In class, I’ve watched students use an LLM on their phone and type the output into a Google Doc on a laptop. I’ve watched students use an LLM and painstakingly write the output by hand into a process journal I asked them to use in class — a process journal that could be presented as future evidence. 🙃
Writing is hard. Trying to make hard, uncomfortable things less hard feels profoundly human to me, so I don’t blame these students. I also believe the more they know about how these tools work, the more empowered they are. We learn about them, talk about them, and, yes, even use them as part of the class. The above examples occurred during a non-graded freewriting exercise for which I asked them to please not use any AI tools. 🙃🙃
I LOVE that you discuss this in class with them. THIS is the education our students deserve. Thank you. 🙏
well when it comes the 45 0 2 Amafia America nq rescue plan
and gotti and online drug distributors
the dea gives them all the covid federal authority
and authorizes construction projects
but the dea doesn’t do that
unless the mafia plan has won
10 2 22 thugaroo 22 once thugaroo and win for the mafia
elon musk can disband
frank 45 0
my boy chops cousin
and so on and so forth
What is an LLM?
Large Language Model. That's what models like Claude, ChatGPT, and Gemini are.
Ah. Ok.
I’m a visual artist and it’s a bit different for us. Some people are advising artists to make all their work on video to prove that it is human made. Videos can be generated by AI too, so I don’t think that’s a long term solution. Many artists have gone back to using physical media instead of digital art. For those using digital art, we are advised to save all of our layers. There is also the issue of AI reference images. AI can generate an image that pulls from another artist’s painting and then the second artist can use the AI reference image and unknowingly plagiarize the original artist. What I’ve been doing is going back to physical media primarily and using only my own reference photographs or ones published online before AI was around. Not to mention that the second my work is put online, it is scraped by AI. It is exhausting and it makes me just want to create artwork in isolation and never put it online.
This is a great article. Well-reasoned, well-written.
As to proving the human provenance of my own work at Forge City Noir, I have taken to "keeping my receipts" in a similar way that you prescribe. When I start a new story, I write out by hand a messy synopsis and/or outline. I put the working title and page number at the top, my name and the date on the bottom of each page.
When I like the outline and it has everything I want to talk about, I start writing the first draft by hand. As with the synopsis/outline, I put the title, page number, my name, and the date on each page.
When I'm ready to transcribe the story, I use a template that has a header and footer to contain all the provenance information, along with the version number. I print that then start marking it up. When I am ready to transcribe the edits, I save off a copy of v1 as v2, change the version in the header, then start applying my changes. Lather. Rinse. Repeat.
I use AI in my process to check historical facts and also as a writing coach. The prompt I developed gives the coach strict instructions to never, under any circumstances, offer me a rewrite. All it's allowed to do is point out an inconsistency, where the story is weak, a plot hole, etc. I'm human and I can't afford an editor, so I use the coach to catch things I miss, but all the work, from conception to final edit, is all mine. And I'm keeping the receipts to prove it.
This is absolutely brilliant. I wish I had your discipline and foresight. Seriously, bravo! And thanks also for your kind words.
Don't forget to cite every source you looked up.
My MC is a young woman. And when I have scenes in which she dresses 👗 up for an event, I describe what she's wearing (because it's first-person POV and she's a little fashion obsessed), which means I have to look up tons of outfits and accessories. Do I need to cite every source? If I Googled Paris Fashion Week, where in my novel do I list the link? Should I include footnotes or endnotes? How long does the reference section of my novel need to be?
This is a very fair comment. I think a lot of the time, I always think about academic writing, whereas you're absolutely right. In reality, every single source for every single idea, of course, does not need to be listed. Also, as the common thread through most of the comments in this piece, I really wish that this kind of provenance wasn't necessary at all and that instead authors were taking it at their word.
The thing that bothers me most about the AI conversation is that it feels like a revenue-driven issue. Really who cares how you made your story? It only matters when money is involved. Corporations don't like it because they feel cheated and worry about their bottom line. Indie authors don't like it for the same reason. Everyone is focused on publishing, not writing. Writers on social media won't talk about writing; they talk about their latest book. The whole issue is about who profits, not who wrote what. Who wrote what is a red herring dressed up like the argument.
I very much appreciate the work you are doing at the forefront of this issue.
I wonder though why we must accept the premise?
Why do these folks wield so much power when they are using such unreliable methods?
In any other field, they would be discredited as folk science.
Very well said, Gustavo. Sadly, I think it's because people in power are not happy about the fact that AI can actually be a great leveller. In many of these cases, it's going to be really interesting to see how this unfolds over the next 12 to 18 months.
It kills me that the onus is put on writers to prove they wrote their own work just because of detection tools that don't work very well, and are getting worse as the models get better at returning the "correct" next word.
Me too Scott, me too. 😢
This is ridiculous. Writers should not have to create a forensic record of their creative process in order to prove they wrote their own work. Drafts are messy. Some of us revise inside the same document. We write on different devices, in notebooks, in fragments, in our heads. We overwrite things. We lose things. That has always been part of writing.
Preserving whatever record naturally exists is sensible. But requiring writers to document their process in anticipation of an unreliable AI detector accusing them turns the presumption completely upside down. The burden should be on anyone making the accusation—not on writers to maintain a chain of custody for their own imagination.
And frankly, my creative process is private. I shouldn’t have to surrender it to prove that my work belongs to me.
I absolutely respect your right for this, Kelly, and honestly I wish this were the case. But sadly, many people are asking for provenance now in a way that just wasn't true even three or four years ago.
And that’s exactly what worries me. I understand that people are asking for provenance now. I’m questioning whether writers should normalize those demands by restructuring and documenting our creative processes around them. There’s a difference between preserving evidence that naturally exists and accepting that we owe anyone a running account of how the work came into being.
I would also like to see the publishers reveal their exact processes...
All developmental editors, copy editors, line editors, ghostwriters, and writing teams need to be named and publicly acknowledged as co-authors. It’s only fair that everyone know who actually wrote the book,
This is the argument I can make it over and over again. Thank you for also putting it into words. AI is just a great leveller, and I would like now all CEOs and alleged creators of work to tell me about their entire process: exactly what they wrote and what was created by ghostwriting instead. Only then can we start talking about AI detection.
I use AI, not in Substack but in other areas. It is an invaluable tool. A TOOL. For one thing I have stout ADD. AI removes a lot of the friction I have to deal with cognitively, such as organizing outbursts of thoughts and finding coherence in scattered observations. Neurotypicals cannot imagine the relief at finally being able to articulate what is swimming around in my head trying to get out.
And I use AI to reclaim my ability to create the artwork for my books. Arthritis has made the amount of work I create quite low. Ai toolset me do things like focus on the figure and let Midjourney or Photoshop fill in dozens of branches or blades of grass. Without that I’d be not creating.
Most of the time they are. That's who authors thank in the acknowledgement pages. But not as co-authors, because (unless they're gravely overstepping their bounds), that's not how editors work. (Ghostwriting and writing teams are a different story.)