The query and pitching process for writers has always been fraught, and the emergence of AI has introduced a new wrinkle. Agents in particular are starting to ask “Was any part of this book or query package created by AI?” in query submittal forms.
I’ve had writers ask me if they have to answer this question yes if they’ve used AI for help in brainstorming, researching, outlining, or editing (but not writing) their materials. But I’ve been caught short without an answer, since I have yet to hear a public discussion among agents on the issue.
So I spoke to several agents and publishers to ask how much evidence of AI use they’re seeing in submitted materials and why they ask about AI use at the query stage. Some were not comfortable going on the record, but everyone I spoke to emphasized that what they value most is the deeply human nature of their work and human writers. And, from a strictly legal perspective, agents and publishers seek assurance that writers are submitting work that qualifies for copyright protection.
Vicky Weber at the Creative Media Agency is one literary agent who asks, via QueryManager, whether writers have used AI. If a writer checks yes, then QueryManager doesn’t let the query go through, and the writer is notified of the block. Weber says she’s had about 25 queries blocked so far because of that question. (The writer can query again and change their answer to the AI question.) Weber says she asks this question because she can’t sell AI-generated material to publishers due to the copyright problem. “It is an automatic no for that reason and that reason alone.”
Weber does not have a problem with writers who use AI for ideation or as a tool to support their own writing process, and such writers can answer no to the AI question. In such cases, she says, “You’re using AI more like you would talk to a friend or your partner to get your own creative ideas flowing. So long as that’s where that ends, that’s different than generation. However, I do think it’s important for writers to keep in mind that AI only can spit out what it’s been fed and what it knows. Sometimes that can mean the ideas that it is giving you are not original. It’s important that you’re using it to spark your own ideas rather than using what it is giving you.”
I asked Mary Laur, my editor at the University of Chicago Press, if they’ve encountered AI use in submissions, and she says editors are now receiving their first book proposals with AI usage statements. One such statement read: “The author acknowledges the use of Claude AI as an editorial and organizational tool in preparing the manuscript. All original content, including the … essays and theoretical framework, was created by the author. At this time, AI assistance was limited to help with manuscript organization and proposal preparation.”
Laur says, “We do not consider this grounds for not pursuing the manuscript, but like everyone else, we’re still figuring out how to deal with all of this.” She added that one of her colleagues feels certain AI has also been used in book proposals that don’t carry any disclaimer.
Ghostwriter Josh Bernoff says if he were evaluating submissions, he’d have two questions for authors. “I would ask ‘Did you use AI to help in research and preparation of this material?’ And ‘Is there any material in here that came directly as an output of an AI tool?’ The person who says yes on the first question should get more consideration … and kicked out on the second question.”
Regardless of whether it’s disclosed, AI use in submissions materials or manuscripts can be painfully obvious to agents and publishers. One agent describes AI as having a BuzzFeed, staccato style, like a “little sass mouth” or clever 22-year-old intern, while another mentions excessive use of adjectives and things coming in threes. (Em dashes did not come up, thankfully.)
Joe Biel, founder of Microcosm Publishing, says his team receives AI-generated materials daily. “We’ve always looked for substance and nuance, which AI is even worse at than humans are. … We ask for and give most consideration to a five-second pitch that distinguishes the book from the rest of the shelf and what readers get out of it. … The immediate tell that someone is using AI is that it says nothing at length and reads worse than what a person would write.” Biel says he’s also had established authors use AI and deny it, but “the editors could tell because the tone and quality of work was so different from most revisions.”
Agent Jennifer Herrera at the David Black Literary Agency worries that AI will only get better over time and that soon its use won’t be obvious at all. Her agency has had discussions about buying into a closed system that produces a score indicating the probability of material having been generated by AI. She believes publishers are already doing this or will be doing it soon because they need confidence the material they’re acquiring or publishing will be protected under copyright.
Still, none of today’s AI detection software is foolproof; it can only deal in probabilities, not certainties. Bernoff thinks AI detection will likely become meaningless as AI improves, plus people will implement myriad customizations that you can’t test for. “Suppose you took the four million words in my blog and programmed a ChatGPT to answer as me. That’s not going to sound like generic ChatGPT output. It’s going to sound like me,” he said. “If you say to one of these [chatbots], I would like you to avoid these words or write this way … that’s going to be very difficult to detect.”
A couple of industry professionals say they currently use AI to help them prepare or improve submissions materials on behalf of their clients. This was particularly true for professionals working on nonfiction book proposals. One person told me it was most useful in helping build out market and marketing plan sections: They can start with an author’s resume or public-facing bio, discuss the author’s assets or expertise, then use AI to help compose or shape that material into sections and pull out themes that work well in a proposal context. Another person told me much the same, saying it was simply smart to do, because AI use in the marketing section or a comparable title analysis will not make a difference to the publisher (assuming you fact-check everything).
Bottom line: Some writers have asked if they should lie when submitting their materials if they think their AI use doesn’t affect the copyrightability of their work. I think that decision can be made only by the writer, who knows their situation best, although I advocate for transparency (no lying). I assume that, over time, clear standards and policies will make it clear when a line has been crossed. One agent admits to me, “It’s a gradient and an obvious ethical transgression at the extreme end, but then it becomes less and less obvious and more and more subtle.”
Bernoff says, “This is going to be a ludicrously irrelevant conversation, because three years from now everybody who writes anything will be using AI as a research helper and ideation helper and to organize things. … The best writers are using this to do what they do better but not to substitute. The mediocre writers, this isn’t helping them that much, because they can’t tell the difference between ChatGPT and what’s good. And poor writers are able to use this to bring them up to mediocre.” He believes writers who don’t use AI at all will end up woefully behind those who do.
The flip side: Can writers trust agents and publishers not to feed their work into AI?
As noted above, some agents and publishers are considering use of AI detection software, similar to what professors use to detect plagiarism in student work. Such software doesn’t train on the uploaded material, nor are those materials shared with AI companies. Materials remain secure and within the local environment of the agency or publishing house.
At the same time, a burgeoning number of startups (e.g., Storywise and Inkbloom) hope to sell overburdened agents and publishers on AI-powered manuscript analysis tools. But Herrera doesn’t see herself using AI tools to evaluate submissions. “The beauty of this business is that we don’t always know what we like. … That’s part of why I remain hopeful that a lot of the tastemaking will be done by humans. People will get sick of formulaic AI stuff.”
There’s also the question of how AI might be helpful to scouts and rights agents who need to generate a considerable number of reports in varying languages for large numbers of titles. Herrera says, “For so long we’ve relied on people to read materials and generate reports for them. … Will AI be doing this?” Her guess is that it’s already happening, but she says that the value of what agents and scouts do “is in the read” and conveying their feelings about a work.
Herrera believes, like I do, that we’ll see more guidelines about how to use AI responsibly. But for now, she believes people are avoiding clarity because the situation is fluid and the landscape might change drastically. “What if [publishers or agents] take a clear stance they have to walk back from?”
Jane Friedman has spent her entire career working in the publishing industry, with a focus on business reporting and author education. Established in 2015, her newsletter The Bottom Line provides nuanced market intelligence to thousands of authors and industry professionals; in 2023, she was named Publishing Commentator of the Year by Digital Book World.
Jane’s expertise regularly features in major media outlets such as The New York Times, The Atlantic, NPR, The Today Show, Wired, The Guardian, Fox News, and BBC. Her book, The Business of Being a Writer, Second Edition (The University of Chicago Press), is used as a classroom text by many writing and publishing degree programs. She reaches thousands through speaking engagements and workshops at diverse venues worldwide, including NYU’s Advanced Publishing Institute, Frankfurt Book Fair, and numerous MFA programs.
