Every few months, someone sends me an article predicting that AI will make human writers obsolete. The articles vary in tone, from breathless excitement to existential dread, but they share a common assumption: that writing is primarily about generating text, and that any technology capable of generating text will eventually replace the people who currently do it for a living.

I think this assumption is wrong, but I also think the opposite assumption, that AI will have no meaningful impact on writers, is equally wrong. The truth, as usual, is complicated and specific. AI will change writing. It is already changing writing. Some of those changes will be genuinely damaging to working writers. Others will be neutral or even helpful. Pretending otherwise, in either direction, does nobody any favors.

I run a publishing house. I work with writers every day. I have used AI tools myself, for various tasks. I want to lay out what I actually see happening, without the hype or the panic.

What AI Can Do Right Now

Let me be specific about capabilities, because vague claims about AI “writing” obscure more than they reveal.

Large language models like GPT-4 and Claude can generate grammatically correct, topically relevant prose on virtually any subject. They can mimic various styles. They can produce first drafts of marketing copy, product descriptions, email newsletters, social media posts, and similar functional text. They can summarize documents, rephrase existing writing, and answer questions in paragraph form.

They can also generate fiction, poetry, and other creative text. Whether that output qualifies as good fiction or good poetry is a separate question that I will address shortly. But the raw capability exists. You can ask an AI to write a short story about a lighthouse keeper dealing with grief, and it will produce something that looks, at a surface level, like a short story about a lighthouse keeper dealing with grief.

What AI cannot do, at least not yet, is think. It does not have experiences, opinions, grudges, obsessions, or the accumulated weight of a life lived. It cannot write from a place of genuine emotional knowledge. It does not know what it feels like to hold your newborn child at 3 a.m. or watch a parent forget your name. It can describe these experiences using patterns learned from millions of texts, but description is not the same as understanding, and readers can usually tell the difference.

The Content Market Is Already Disrupted

If you write commodity content for a living, you are in trouble. I say this with sympathy, because I know many talented people who pay their bills by writing blog posts, product descriptions, press releases, and similar material for businesses. This segment of the writing market has already been significantly disrupted, and the disruption will continue.

The economics are straightforward. If a business was paying a freelancer $200 to write a blog post, and an AI can produce a passable version of that post in thirty seconds for essentially zero marginal cost, the business will use the AI. Not always. Not for everything. But for a large enough portion of the market that many content writers have already seen their income decline.

I have talked to content writers who have lost 40-60% of their freelance income over the past two years. These are not bad writers. Some of them are very good. But the work they were doing, informational blog posts, SEO articles, standard marketing copy, has been automated to a degree that is hard to compete with on price.

This is a real harm affecting real people right now, and I get frustrated when AI boosters dismiss it as creative destruction or suggest that affected writers should simply “move up the value chain.” Moving up the value chain requires time, opportunity, and financial runway. Not everyone has those resources.

At the same time, I have noticed something interesting. The businesses that replaced human writers with AI-generated content are, in some cases, starting to notice a quality gap. The AI output is fine. It is correct. It is adequate. But it is also flat and interchangeable. It reads like everything else on the internet because, in a sense, it was trained on everything else on the internet. Some of those businesses are coming back to human writers, willing to pay more for work that has a distinctive voice and genuine expertise behind it.

This does not help the writers who lost income in the interim. But it suggests that the market may settle into a new equilibrium where AI handles the commodity layer and human writers handle the premium layer. The premium layer will be smaller, but it may pay better per piece.

What About Books?

Book-length writing is a different animal from content marketing, and the impact of AI here is more nuanced.

Can AI write a novel? Technically, yes. People have published AI-generated novels. Amazon had to implement restrictions on AI-generated book uploads because the marketplace was being flooded with them. Some of these books are transparently terrible. Others are surprisingly readable in the way that a fast-food burger is surprisingly edible: it satisfies the most basic need, but you would not confuse it with a meal cooked by someone who cares.

The novels that readers remember, recommend to friends, and carry with them for years are not produced by assembling statistically likely word sequences. They come from writers who have something specific to say and a specific way of saying it. When Catherine Voss wrote The Last Archive, she brought two decades of thinking about memory and institutional power to the page. An AI could produce a thriller about archives. It could not produce that thriller, because that thriller required Catherine Voss to exist and to have lived the particular life she has lived.

This is the argument I find most persuasive about literature’s durability in the face of AI: great writing is not just about the text. It is about the mind behind the text. Readers form relationships with authors. They follow their work, attend their readings, care about their opinions. An AI does not have a mind to form a relationship with. It does not evolve over a career, struggle through a difficult second novel, or surprise everyone with an unexpected genre shift at age 55.

But I want to be honest about the risks too. The flood of cheap AI-generated books on retail platforms is a real problem. It makes discovery harder for human authors. It depresses prices. It fills search results with noise. And for genres where formula matters more than voice, like certain categories of romance, thriller, and science fiction, AI-generated books can be “good enough” for a segment of readers who are primarily looking for plot delivery rather than literary distinction.

AI as a Writing Tool

Here is where the conversation gets more interesting to me, because this is where the practical reality of AI’s impact on professional writers actually lives.

Many writers, including some excellent ones, have started using AI as a tool in their process. Not to write their books for them, but to assist with specific tasks along the way. Research. Brainstorming. Outlining. Checking facts. Generating placeholder text that they then rewrite entirely. Asking “what am I missing?” after finishing a draft. Using it as a sounding board when they are stuck.

I know a nonfiction author who uses AI to help organize research notes. She feeds in dozens of sources and asks the tool to identify themes and contradictions. She then verifies everything independently and does all the actual writing herself. But the organizational step, which used to take her weeks, now takes hours. She writes better books faster, and the books are entirely her own.

I know a novelist who uses AI to generate character backstories that he then discards almost entirely, but the process of reading and rejecting AI-generated ideas helps him figure out what he actually wants. It is a creative sparring partner, a way of thinking out loud.

These uses are not threatening. They are productivity enhancements that leave the creative decisions firmly in human hands. They are no more “cheating” than using a word processor instead of a typewriter, or using Google instead of a physical encyclopedia.

The line gets blurrier with ghostwriting and collaborative writing. If a nonfiction expert provides their ideas and expertise, and an AI generates a first draft that a human editor then substantially revises, is the resulting book “AI-generated”? I would argue no, at least not in any meaningful sense. The intellectual content came from the human. The writing was shaped and refined by humans. The AI was a tool in the middle, like a very fast and very mediocre research assistant.

But reasonable people disagree about this, and the publishing industry has not yet developed clear norms or disclosure requirements. At ScrollWorks, our policy is straightforward: we publish books written by humans. Our authors may use AI tools to assist with research, organization, or brainstorming, but the writing itself must be substantially theirs. We ask about this directly, and we trust our authors to be honest. So far, this has not been a problem.

The Economics of the Transition

Let me talk about money, because ultimately that is what determines whether writers can continue to exist as professionals.

The writing profession was already under severe economic pressure before AI arrived. Advances have stagnated or declined in real terms. Freelance rates for journalism and content writing have been flat for over a decade. Many working writers earn below minimum wage when you calculate their hourly income. The Authors Guild’s last income survey found that the median income from writing for full-time authors was around $20,000, which is below the federal poverty line for a family.

AI is adding pressure to an already pressured system. The writers most affected are those who were already in the most precarious positions: freelancers, content writers, early-career authors without established reputations. The writers least affected, so far, are those with strong personal brands, dedicated readerships, and the ability to do things AI cannot replicate.

This dynamic risks creating an even more stratified writing profession. A small number of successful authors will do well, perhaps even better than before, because AI tools may boost their productivity. A larger number of aspiring and mid-career writers will find fewer opportunities and lower pay. The pipeline that turns aspiring writers into established ones may narrow significantly.

This is the scenario that worries me most. Not that AI will replace the best writers, but that it will eliminate the early-career opportunities that allow writers to develop into their best selves. If you cannot make a living writing blog posts and magazine articles while you work on your novel, you might never write the novel. And we will all be poorer for it.

Copyright and Training Data

The legal questions around AI and writing are far from settled, and they matter enormously to working writers.

Current AI models were trained on vast quantities of text scraped from the internet, including copyrighted books, articles, blog posts, and other writing. Several lawsuits are currently working through the courts, including actions by the Authors Guild and individual authors against AI companies. The core question is whether using copyrighted material to train AI models constitutes fair use or infringement.

I am not a lawyer, so I will not predict how these cases will be resolved. But I will offer an opinion: writers should be compensated when their work is used to train commercial AI systems. The fact that the training process is technically complex does not change the fundamental principle. If someone profits from your work, you should receive something in return. The specifics of how to implement this (licensing fees, collective bargaining, opt-in versus opt-out) are details to be worked out, but the principle seems clear.

Some AI companies have started making licensing deals with publishers and media organizations. This is a step in the right direction, but the deals so far have mostly involved large publishers and news organizations. Individual authors and small publishers like ScrollWorks have limited leverage in these negotiations. Industry organizations will need to play a role in ensuring that compensation reaches the writers whose work actually fuels these systems.

What Writers Should Actually Do

I promised an honest assessment, so here is my honest advice to writers navigating this landscape.

If you write commodity content, diversify now. Do not wait for the market to finish shifting. Build expertise in areas where your personal knowledge and experience add value that AI cannot replicate. Develop a recognizable voice. Position yourself as a specialist rather than a generalist.

If you write books, focus on what makes your work uniquely yours. The books that will thrive in an AI-saturated market are books with strong authorial voice, genuine expertise, and the kind of specificity that comes from lived experience. The Cartographer’s Dilemma works because David Okonkwo brings his specific background in geography and West African history to the narrative. No AI could write that book, because no AI has David’s particular combination of knowledge and perspective.

Learn to use AI tools where they genuinely help. Do not be a Luddite about this. If an AI tool can save you three hours of research organization so you have more time for actual writing, that is a good trade. The goal is to use technology to amplify your human strengths, not to replace them.

Support collective action. Join the Authors Guild or similar organizations. The challenges AI poses to writers are structural and cannot be addressed by individual effort alone. Copyright protection, fair compensation for training data, platform regulation, and professional standards all require organized advocacy.

And keep writing. I know this sounds trite, but it is the most practical advice I can give. The writers who will emerge from this transition in the strongest position are the ones who continue to develop their craft, build their audience, and produce work that only they can produce. AI raises the floor. It generates competent, unremarkable text with unprecedented ease. But it does not raise the ceiling. The best human writing is as far above AI output as it has ever been above mediocre human output. If you are working to raise your ceiling, you will be fine.

What Publishers Should Do

Since I run a publishing house, I should say something about our side of this.

Publishers need clear, enforceable AI policies. Authors need to know what is expected of them, and readers need to know what they are buying. Transparency is non-negotiable. If a book was substantially written by AI, readers have a right to know. If it was written by a human with AI assistance for research and brainstorming, that is a different situation, but norms around disclosure are still evolving and publishers should be on the leading edge, not lagging behind.

Publishers also need to invest in their authors more, not less, during this transition. Author platform development, marketing support, editorial engagement: these are the things that differentiate a human-authored book from an AI-generated one. If a publisher’s value proposition is just “we print and distribute text,” then yes, AI threatens that business. But if the value proposition is “we develop and champion distinctive voices,” then AI actually strengthens the case for what publishers do.

At ScrollWorks, we are investing more in editorial development than ever before. We are spending more time with our authors on structural editing, voice refinement, and the kind of close attention that turns a good manuscript into an excellent book. We believe that the gap between AI-generated text and carefully edited, human-authored work will only become more obvious over time, and we want our books to be on the right side of that gap.

The Long View

I will close with a perspective that I think gets lost in the day-to-day anxiety about AI and writing.

Humans have been telling stories for at least 40,000 years. We told them before we had written language. We told them before we had paper, printing presses, typewriters, word processors, or the internet. Every new technology has changed how we write, but none has eliminated the human need to hear another human being say: here is what happened to me, here is what I saw, here is what I think it means.

AI will change the economics of writing. It will change the market for certain kinds of text. It will force writers to adapt, and some of those adaptations will be painful. But it will not replace the human impulse to write, or the human desire to read something written by another person who has something genuine to say. That desire is older than civilization, and no algorithm will extinguish it.

The honest assessment, the one I promised in the title, is this: AI is a serious challenge for working writers, particularly in the short and medium term. It is not an existential threat to writing itself. The writers who adapt will survive and possibly thrive. The ones who do not adapt will struggle. And the industry, publishers, agents, booksellers, and libraries, has a responsibility to help writers through this transition rather than leaving them to figure it out alone.

We owe writers at least that much.

The ScrollWorks Media editorial team publishes books by human authors. Explore our catalog to find writing worth your time.