Creation vs. Generation: Drawing a Non-technical Red Line Between Human Authorship and AI

True authorship is a uniquely human enterprise centered on a bond between writer and reader. To protect this relationship, writers must draw a bright red line between deterministic assistive tools (like spell check, thesaurus, and transcription) and "generative AI" (which adds, rearranges, or rephrases words). This generative AI dilutes voice, lacks an understanding of meaning, and abdicates authorial responsibility, making it an existential threat to genuine creativity.

AI is insidious. Driven by profit and hype, it's showing up anywhere companies can shove it. The recent encyclical Magnifica Humanitus offers a profound take on this topic, concluding that artificial intelligence actively displaces human creativity. Authorship, in particular, is a human enterprise that centers on bonding readers to writers. To suggest that AI could replace, augment, or facilitate that bond severely misunderstands what creation means. Bardsy's position on AI is clear: AI threatens creativity. We are an AI-Free Zone for this reason and do not countenance the use of generative AI. For example, our anthology contest entrants must affirm that they do not use this shortcut. Writers who can't make this pledge have their entry fees refunded. Understandably, however, this seemingly simple stance has led to questions.

Similar confusion attends recent author contracts. Every one we've reviewed in the past year contains a blanket clause prohibiting the use of AI in the work's production. While we applaud every action publishers take to limit AI's infiltration, these clauses lack specificity and deny responsibility. They are, legally speaking, appallingly vague and practically unenforceable. Is the clause violated when a Nest thermostat, for instance, regulates your office's temperature? What about using a thesaurus? The answers are almost certainly no. We must also see businesses' tendency to claim their product has AI when it really doesn't. These contracts leave authors vulnerable. Further, the publishers behind them are abdicating responsibility for defining AI and transferring the burden for its limitation to authors.

More broadly, this confusion and vulnerability stem from the failure to define AI in practical terms that make sense to nontechnical people. This post's purpose is not to separate good AI from bad. Instead, we seek to define AI, so it can be avoided. In other words, we want to draw a bright red line between AI and humanity. Staying on the right side, then, makes you part of the solution and ensures that your Bardsy anthology contest entry is good to go.

Critically, our line eschews complexity in favor of looking at what authors do. Put simply, we want to remove AI from that process. Defining AI this way eliminates technical discussion. Meanwhile, it allows the use of assistive technology to support your efforts. We're going to start from our framework for creating compelling novels, specifically the definitions of story, novel, and writing. You don't have to be an expert to follow these rules and stay on the right side of this battle.

There is a lot of pressure coming from technology companies to simply cut humans, including authors, out of the loop. Why? To increase profits. From a corporate perspective, what could be better than manufacturing books without having to deal with annoying, unpredictable and possibly expensive talent?

Fighting back calls for labeling laws and boycotting AI work. You can do more. Writers must understand their value and use that knowledge to purge AI from their process and educate fans. The bottom line for authors is that people want to read what other humans write. To do that, we must move past technical descriptions of language models, which most of us don't care about, and develop a concrete list of what to avoid and why.

The surest route to clarity in battling AI lies in defining authorship. Bardsy offers this framework:

Story: A coherent combination of five elements: character, world, conflict, theme and plot.
Novel: A story effectively expressed in words, typically 90,000.
Writing: The process of perfecting the story in your mind and putting it into words for readers.

Writing itself, then, involves three tasks:

Deciding what to say and how to say it. Choosing words when formulating sentences generally happens at the paragraph level while broader story architecture tends to be less specific.

Deciding what information the reader needs and delivering it at every point in the story so it appears in an inviting order.

Allocating words so that some points land with greater impact than others, helping readers experience stories as you intend.

It may require some reflection to see these three items as separate tasks because they overlap and often happen outside of our self-awareness. Nevertheless, this set fully characterizes authorship. In doing so, this framework captures the value your efforts add.

This background points out what technology to adopt and what to avoid. We often, for example, get asked if spell check counts as AI. The answer, of course, is no. Checkers are an assistive technology and, like their cousins targeting grammar and format, predate the AI craze. These tools are fine. You wouldn't prevent authors from looking up a word in a dictionary or having a human proofread punctuation. There's no additional onus because computers mediate this task. Moreover, calling them AI is misleading because they are deterministic.

What about transcription services? In the old days, professors (and other poor typists) used to dictate pages for others to type. Today, some authors avoid keyboards in favor of assistive technology, which allows them to use alternative inputs. This is not AI, no matter what some companies claim. Why? Because authorship is intact. We can trace the origin of the text directly to the author's mind. With this provenance assured, nothing further needs to be said.

What about using a thesaurus? Thesauri have been writers' companions at least since Roget published in 1852. More to the point, a thesaurus doesn't "generate" words; instead, it offers close to one-to-one substitutions. The author must review the list and choose the best. Naturally, you can choose poorly. To generalize: drawing this line looks at non-random application, provenance, and the substance added.

A straightforward test imagines two windows on your writing device. One has your text while the other offers some version of AI, be it Copilot, Opus, or whatever comes along. We encourage you to see this other window as an ongoing temptation, fine to use for a spell check, transcription, or thesaurus, but no more.

To keep your work clean, follow these three rules:

The meaning must come from the author's brain. Seeing the word count needle move up without real effort is a massive red flag. Whether it's a single phrase or whole pages, those are not your words. The most obvious transgression arises when a prompt produces more new words than it offers. A six-word prompt, for instance, like "write a conversation between two main characters" leads to paragraphs of output. You can't create something from nothing, so the extra words have no provenance. Chances are they've been plagiarized in a diluted and random form from other authors' works. This must be avoided at all costs.

Some AI products claim they can write a chapter for you or, worse, write a novel from your outline. Look closely and you'll see their output is, once again, largely random. If, for instance, you criticize one's output, it will not learn and provide a better order. It will only provide another semi-random version. This deception is a fortune-teller's trick of spinning options until one sticks.

Danger arises every time you cede creativity or control. Authors must empathize with readers in deciding how to present their story. When you give that task to a machine, you also break the authorial contract. Readers can't trust that an artwork has meaning when a machine organizes the experience.

This is the most pervasive contemporary use of AI. Many writers turn to machines to "fix" sentences they can't quite get right. This technology qualitatively differs from thesauri. For example, Grammarly loves to suggest alternative phrasings because it's easy for its software to roll the dice, which combines with a vast thesaurus to offer unlimited variations. The conjunction of randomness and options changes the provenance. Further, using this machinery subtracts value. It destroys your voice and reduces the probability that the reader will experience the story as you intend. It is the exact opposite of what crafting a novel is all about.

The line drawn above leads to a better understanding of why AI cannot, even in part, take the place of human authors.

AI has no understanding of what it is saying. It is just arbitrary data arranged to look like coherent text through its training. A compelling novel, in contrast, transfers meaning from author to reader and produces shared understanding.

Think of your story as a pristine pond. Introducing AI is like dumping garbage into its clean water. The pollution obscures your intent rather than furthering it. Successful novels, on the other hand, depend on effectively transmitting a coherent story, an author's singular vision.

There was a competition for AI-written stories recently; the winner was terrible. AI loses "concentration" (if we can call it that) over longer formats. It can't produce a coherent chapter, let alone put continuing chapters, with ongoing themes and arcs, together for 100,000 words.

You wouldn't trust an AI doctor to prescribe medicine because an AI bears no responsibility. A book represents an almost contractual, albeit parasocial, relationship between two human beings. Readers buy books because they know another human wrote them. AI can never hold up its end of that bargain.

Finally, we must appreciate the difference between two key words, starting with generation. Generation makes us think of a power plant, like a machine where inputs and outputs are proportional. An AI "works," i.e. produces garbage, only because it has chewed through the millions of human words as well as your prompt. It adds nothing.

Creativity demands disproportional inputs and outputs; your brain is adding value to the new whole. Authors, thus, synergistically enhance their experiences to create something new for real readers. To be a true creative, stay on the right side of the red line. Educate those who cross it. There is some good news: over the course of this struggle effective action on the part of AI's foes could produce a healthier, more human society.