All Resources

By Sandi Sonnenfeld

This article originally appeared in the Summer/Fall 2022 issue of the Authors Guild Bulletin.

Humanity has long had a love-hate relationship with technology. That’s because despite the benefits new technologies bring, they also nearly always result in unforeseen, and often harmful, consequences.

The cotton gin’s creation in 1793 revolutionized the textile industry by making it possible to separate cotton fibers from their seeds more quickly. It also significantly increased the demand for cotton workers in the American South, adding to the growth of slavery. The gas-powered automobile, and later the airplane, made traveling great distances easier, faster, and cheaper, fostering a better understanding of differing cultures. It also led to the building of highway systems that displaced more than one million Americans, separated or isolated poorer neighborhoods, destroyed acres of greenbelts, worsened air quality, and helped accelerate climate change.

Few professional writers today need to be told about both the immense benefits and the adverse impact that the Internet Age and the companies and technologies that drive it, Amazon and Kindle Publishing Direct in particular, has had on the creation, production, distribution, and sales of books as well as the steady decline in the annual median income of book authors.

Enter AI

Perhaps no technology, however, has both inspired the imagination and raised more concerns than artificial intelligence—machines that serve, function, sound, behave, and look like humans, in the form of computers, robots, and androids. One has only to think of the monster in Mary Shelley’s Frankenstein or the sentient computer HAL who runs amok in Arthur C. Clarke’s 2001: A Space Odyssey, to understand why writers remain both fascinated and horrified by AI, particularly given the irony that human beings are responsible for creating the very instruments of their downfall.

Because we’ve already integrated some AI-based technology into our daily lives—as evidenced by our reliance on smartphones, personal assistance devices like Alexa and Siri, and wearables like Apple Watches, Fitbits, virtual- and augmented-reality headsets and glasses, AI hearing aids, and biometric devices—many people remain unconcerned about the future of AI.

Indeed, thousands of writers already deploy software and apps that incorporate some AI elements. Grammarly helps edit one’s prose and even offers alternative wordings of sentences to improve clarity and flow. Scrivener serves as a novel outline and chapter planning tool, while Dramatica helps identify information, plot points, or character interplay that writers might be missing to flesh out their text and resonate more with readers. None of these tools from recent years will actually write the novel, story, or screenplay for you; they only help. But that is starting to change because of rapid strides in machine deep learning, in particular the creation of GPT-3, a text generator that can spit out a movie script in a matter of minutes, and DALL·E 2, an image generator that can conjure up visuals based on any combination of words.

Both GPT-3 and DALL·E 2 are from the research lab OpenAI, in which, it should be noted, Elon Musk, Amazon, and Microsoft are huge investors. GPT-3 belongs to a category of deep learning known as a large language model (LLM), a complex neural net that has been “trained” to write by consuming a ginormous amount of text from digitized books from around the web. GPT-3 is the most celebrated of the large language models, and the most publicly available, but Google, Meta, and DeepMind have all developed their own LLMs in recent years.

Below are some AI writing platforms that use GPT-3 technology to produce novels and stories with limited input from humans.


Sudowrite, which allows users to try it free before committing to either a monthly $29 professional subscription of 50,000 words per month or a $9.99 “student and hobbyist” offering of 10,000 words per month,7 appears to be the best reviewed among the new AI writing platforms. A fiction writer types an original first sentence of a short story or novel into the app, and Sudowrite then crafts an opening paragraph from that sentence. If the writer isn’t happy with the text, she can then instruct Sudowrite to “add more description” or “show not tell” and the program will respond to the directions given. It then builds the next paragraph based on the first one and goes on from there.

“While the story isn’t Pulitzer Prize–worthy (yet), I was startled by the algorithm’s ability to turn phrases like ‘the tall grass swayed in an unseen breeze’ and ‘its hushed breathing carried an echo of a crackling campfire,’” short story writer Nick Bilton recently wrote in Vanity Fair after test-driving Sudowrite.


Originally conceived for use by advertising and content marketing agencies or companies with marketing communications needs, Jasper offers 50-plus templates in 25 languages to write SEO-focused blog posts, web content, ad copy, listicles, journalism articles, school papers, video scripts, and books.

While it offers a basic subscription starting at $29 a month, most users opt for the “Boss Mode” subscription at $119 a month. One reviewer noted of Jasper, “While it isn’t the cheapest AI writer on the market, the competitive features, clean user interface and high quality output potential maintains it as great value for folks looking to do a large amount of writing and copywriting without the need to hire expensive outside help.”


NovelAI is a monthly subscription service for AI-assisted authorship, storytelling, or virtual companionship. The company promises that by using AI algorithms to create humanlike writing based on one’s own voice, anyone—regardless of ability— can produce quality literature.


ShortlyAI is a powerful piece of AI novel-writing software that can help users get past writer’s block, organize thoughts into well-written paragraphs, brainstorm, and more. It even claims that it helps users write 10 times faster. The software, which is priced at $780 annually or $79 per month, features a set of commands for refining sentences in a variety of ways, including rewriting, shortening, and expanding content.

Regardless of the platform users select, however, OpenAI’s GPT-3 program is not without limitations:

  • It does not allow users to generate “adult” or “sexual” content due to the service terms.
  • The AI can only “look back” so far in the text to connect all ideas together and ensure a cohesive storyline and uniform syntax. That means it can’t write an entire book in a single session, and users will still need to “remind” the AI of major plotlines, character names, and so on.
  • Users must generate all text in the chosen platform’s native editor, so one cannot use it with Microsoft Word, Google Docs, or another word processing program. This raises some questions about the final output and sharing of the document in process with an agent, editor, or publisher. However, most platforms have built-in grammar-checking programs that help with the editing process.

Platform reviewers also mention that some of the texts generated may come across as biased or stereotyped because the AI is drawing on millions of sentences and paragraphs written over the many decades when white, straight males dominated the writing landscape.

Sodowrite, Jasper, NovelAI, and ShortlyAI all still require human beings to review the prose, suggest changes, ensure the absence of offensive language or tone, and confirm accuracy and meaning. Because AI technology can continuously learn, however, many of these problems will eventually be eradicated.

It is only a matter of time before some Hollywood executives, internet platforms, online publications, and even publishers opt to eliminate the human writer from the equation (aside, perhaps, for edits) and use Sudowrite or other AI platforms to produce manuscripts. Jasper already lists HarperCollins as a client, though it’s unclear what purpose the publisher uses the platform for.

Economic impact of AI on writers

In 2017, McKinsey & Company predicted that AI technology could replace up to 42 percent of workers in the arts and entertainment sector by as early as 2035, including book authors, screen and television writers, and playwrights.

That’s a terrifying proposition given that most American authors today earn so little that writing is already becoming an unsustainable profession. A 2019 Authors Guild survey found that the median writing income for full-time authors was just $20,300. With the development of AI platforms like Sodowrite and Jasper, publishers might be able to produce books even more cheaply, genre fiction in particular, which already relies on formulaic elements. Once the cost of a sophisticated AI-writing machine is amortized, the cost of having it write and create other works will be nearing zero. Human writers will not be able to compete with that.

Moreover, if publishers decide to publish AI-created books rather than buy manuscripts from human authors and pay both advances and royalties, human authors who continue to publish are likely to see the amount of their advances shrink. Writers need advances both to do the necessary research and to live on while writing their books. AI machines can not only write books more quickly than humans, if the hype can be believed; they also needn’t worry about money for food, rent, utilities, health care, or clothing.


Even as late as 2014, when the Authors Guild’s lawsuit against Google for its book scanning project was decided in the company’s favor (ignoring the huge value that Google received from the mass number of book scans it used to teach its computers to write), the idea that human creativity could be co-opted by computers still seemed more a nightmare from a sci-fi novel than reality. Since then, deep learning has accelerated so rapidly that AI technologies now can so closely mirror the expression of human emotions and thought that it is becoming increasingly more difficult to differentiate between literary material written by a person and that by an AI machine.

Indeed, a recent Washington Post article about how Google had put one of the ethicists working in its Responsible AI group on administrative leave after he went public with his concerns that Google’s AI machine is sentient ended on a terrifying note: “Google says there is so much data, AI doesn’t need to be sentient to feel real.”