Tag: chatgpt

Generative AI will steal your job if you let it. By that, I mean that you’ll be drowned out by boring, confusing garbage if you do not develop your own unique voice.

I do not think that people will ever stop wanting original fiction, but as generative AI strengthens its capabilities, indiscriminate readers will be unable to differentiate the middling author from the stupid plagiarism machine.

So you need to create a unique writing style, which can be pretty challenging. The first step is to read widely and incorporate anything you like into your writing for a while, like a literary hermit crab. Then, as you continue to grow in confidence and capability, you’ll learn how to blend these together into something wholly your own.

What I provide here is simply a guide about what AI does regularly, not a bible for developing your voice. Hopefully, this can help you understand how to outcompete AI in fiction writing so that people want to read your work, not generative slop.

Write longer pieces.

Most generative AI has a problem with maintaining a consistent plot over more than a few thousand words. These programs have short attention spans and will invariably begin to repeat themselves.

While some clever AI “writers” will double-check their “work” and ensure everything flows naturally, a lot of others will ignore these problems and publish the slop just as it is. This, of course, annoys readers who have a longer attention span than a gnat.

Writing a longer piece with AI is honestly more work than most lazy “writers” want to do because they’ll have to prompt the program over and over again, reminding it of the plot at regular intervals. So, you’re in luck if you actually like writing, because you can easily create a 100+ page work without constantly repeating yourself.

Include controversial content.

If you’re writing for an adult audience, it is okay – encouraged, even – to include violence, drugs, explicatives, and sex. Don’t have these in children’s books, obviously.

AI is curated to satisfy advertisers, not the general public, so they have strict moderation to avoid certain topics. If you ask AI to give you a bomb-making recipe, it won’t. Request a graphic torture scene, and you’ll get a hard no. Even things like contentious political arguments will get shut down because, again, advertisers don’t want to court controversy.

So, AI content is generally fluffy, happy-go-lucky bullshit. Any conflict gets solved through the power of love and friendship, and any difficult subjects are endlessly justified or contextualized for the audience like we’re all idiots.

Simply think about what would piss off a big company and you’re likely to have something good on your hands. A few swear words here and there can make a difference in whether your work reads like AI or not.

Of course, you don’t have to include these things if you don’t want to. However, writing thoughtfully about icky topics will demonstrate, without a shadow of a doubt, that a human was at least partially responsible for this piece’s creation.

Develop unusual plot twists.

Always remember that AI does not think like a human; it cannot create unique plot twists because it just follows standard formats. As such, things we would consider a natural if unexpected turn would never occur to AI.

For example, if the so-called “writer” tells AI to write a story about heroes defeating dragons, it would never do something like, say, the heroes defeating the dragon by luring it out during a pounding rainstorm so it can’t use fireball attacks, then using a crowbar to pry off a few scales.

That, to AI, would be too absurd to fathom, even though humans could easily see that happening. Put your creativity to work, identifying ways to seamlessly connect different scenes into an overarching narrative.

Make logical leaps.

Because AI cannot think like a person, it cannot spontaneously connect two disparate concepts to make a novel thought. You see this often in third-person limited, where the “inner thoughts” will just be the most surface-level conclusion.

For literature, these logical leaps will often come in the form of unusual metaphors or similes. One of my favorite examples that I’ve written came from a fanfic I wrote, where the MC compares a burning castle to the decomposition process:

There is something organic about the way this place decays, like watching an animal decompose. First it puckers at the edifice, turning ash grey instead of bone white; its teal roof tiles blister and pop and shed. The gold ornaments on its eaves seem to protrude further out as the wood holding them slivers away, and Ichigo remembers Yagen telling him that people used to believe that hair and nails grow after death because the skin shrinks back. So like a living being, this castle – the changes that come to all things as they fade away.

This is part of the reason that writers need to read nonfiction as well as fiction. By gathering unique factoids, you can make logical leaps between unrelated ideas to create interesting parallels.

Avoid cliches (like the plague).

Because generative AI has been trained on thousands of books, most of which include the same cliches, it relentlessly regurgitates these throughout any of its responses. As such, avoiding cliched terms like “his heart dropped” or “his throat tightened” can make your work more interesting and less robotic.

Utilize subtext.

AI cannot do subtext. Everyone alway says exactly what they mean. There are no subtle gestures that overshadow a character’s spoken words, no hidden symbolism in clothing or room layout, no deep themes that are quietly repeated throughout a text.

You can outpace AI by simply thinking harder (because AI doesn’t think at all). Figure out what different elements of your story really mean; allow characters to contradict themselves in action versus speech.

Craft interesting descriptions.

Because AI does not imagine things, it cannot visualize new settings or get too precise with its descriptions. So, no matter what TikTok girlies may think, description is truly the key to success in a post-AI world.

Get intensely visual. Point out unique items in a room that have hidden meanings, or rhapsodize about the sunset over the prairie. Bonus points if these all contribute to the overall symbolism and subtly impact the overall mood.

Cut out any purple prose.

AI loves to go way overboard with the melodrama when you tell it to write fiction. It will provide long, weird, dramatic sentences that read like they were lifted from a hideous Victorian flop.

As I’ve said before, elegance isn’t automatically purple prose. If you balance short, straightforward sentence with more flowing, symbolic ones, you can keep your work from sounding maudlin like AI.

Remove double explanations.

When writing fiction, AI tends to use very similar sentence structures every time. This is especially noticeable in the fact that it will have an independent clause, then a dependant clause that essentially repeat the premise of the independent clause.

For example, it might say “She smiled with ruthless intent, a sign of her cold and contemptuous nature.” We really didn’t need the dependent clause because it doesn’t offer anything new.

AI always has similar sentence lengths, too, usually about a line and a half. This makes the story feel boring and repetitive, even when it’s not repeating itself.

Using different sentence structures and lengths will spice up your writing, putting you leagues ahead of AI.

Create ambiguous endings without a clear “wrap-up.”

AI loves to wrap a story up with some moral or overarching theme, probably because it was trained on thousands of stupid essays where the writer had to summarize everything at the end.

Remember that AI was primarily trained not to do creative writing, but to summarize and explain things to an audience, like what you see with Google’s AI search function. As such, it will always try to do this even when the format doesn’t require it.

You do not need this. Readers can figure out for themselves what your story meant. Keeping things ambiguous makes your reader figure things out for themselves, which AI does not want them to do.

Break writing rules – within reason.

As I’ve mentioned previously, writing rules are not set in stone. They are training wheels that are meant to help you develop good writing habits while leaving room for experimentation.

Artificial Intelligence does not understand this, so it will ruthlessly adhere to common writing standards. While this makes for technically proficient prose, it also makes AI writing sound wooden and lame.

So play around a little. Add a small section of present tense prose into a primarily past-tense novel. Do a run-on sentence, or even a run-on paragraph. Insert different writing formats, like an email or meeting notes, to summarize something.

Contained selections of rule-breaking prose will not usually annoy your reader; it’s when you do an entire gimmicky piece where people click off. Create a good balance between “normal,” standard writing and something a bit more experimental.

Blend writing eras and styles.

This one is more challenging, which is why I have left it for last. Basically, AI will only write in one certain style depending on what you ask. It cannot and will not play around with different tones.

I developed my unique style by dint of being an American who read a lot of British literature as a child and young adult. My voice is a mix of modern and Victorian, which makes it stand out in today’s world. I phrase things differently from both American and British authors since I was surrounded by stuffy pretentious manor novels at a formative age.

This tonal shift is subtle and difficult to articulate, but it’s undoubtly present. Some of this comes down to terminology, like “succor,” which was very common in Victorian-age and earlier Brit lit. In fact, since I learned this term from novels, I didn’t even realize that the American spelling doesn’t include the “u.”

Other things that are apparent in my novels are elements like the word “mustn’t,” or the phrase “would that I,” and so on. These work well for fantasy novels, where it’s expected that people will speak in a more old-world style.

I’m also a business and SEO writer with an International Relations degree, so my work often has a more formal tone than one might see in other fantasy novels. That’s what I’m accustomed to, and it bleeds through into my fiction.

But AI doesn’t do that. It cannot adopt a formal tone in, say, a fantasy novel; it will go off into this melodramatic purple prose. Similarly, it won’t use a freewheeling, casual voice for a thriller because to AI, that’s just not what’s done.

To mix eras and writing styles, read widely, as I mentioned before. If you have been reading modern literature, focus on older lit. If you mostly read American or European stuff, branch out to Asian or Middle Eastern translations. There will be so many differences that you’ll start to pick up, like terms of endearment, phraseology, and so on.

Pessisimistic as many people are about the future of writing, I’m not particularly concerned. AI may please surface-level readers, like TikTok girlies who skip half the book anyway, but it can never satisfy the deep human urges that make reading so fun.

Don’t get it twisted: things will be uncomfortable for a while. We can’t discount the terrible effects that artificial intelligence has on the real world, from accelerating climate change to damaging children’s critical thinking skills. Publishers are turning to cheap, lazy AI writing because they can’t think beyond the short term and don’t really care about good products.

These are real, present dangers, both to the environment and to society as a whole. But they aren’t the scary, literature-ruining boogeyman that people are so worried about. AI isn’t going to destroy book publishing forever.

Eventually, AI will pass like every other writing fad, whether that’s pump-and-dump genre fiction or choose-your-own-adventure books for kids. Those of us who care about our craft will remain, providing fresh and thoughtful content that stands the test of time.