How to Get ChatGPT to Write Better Fiction

As an AI, ChatGPT does not have any innate creativity. It cannot come up with original concepts on its own or use its imagination. It needs to be given specific requests (called prompts) in order to generate bodies of text.

Keep in mind that ChatGPT remembers what you said in earlier parts of a given conversation and that you can correct it with feedback. As for its limitations, the most important one for getting it to write fiction is that ChatGPT needs to use existing world information and prompts; it can’t come up with anything original by itself.

We’ll teach you how to get ChatGPT to generate more interesting content, how to enhance its style, and how machine learning works so that you can further experiment on your own.

However, first, we need to break down the most important elements of fiction writing and how ChatGPT uses these elements.

There’s a lot that goes into creating an interesting and engaging story. This can make it quite tricky to generate useful prompts. That’s why it’s crucial to phrase your prompts properly to give ChatGPT the best possible chances of generating worthwhile text.

The sequence of events that make up the story. This includes the event that begins your tale, the factors that push the story and characters along a certain path, the climactic moment, the aftereffects of that climactic moment, and some sort of resolution.

ChatGPT has been trained through many texts—including fiction ones. When asked to generate fiction, it tries to replicate patterns found in existing works of fiction. It will always use the following to generate a plot:

• Exposition
• Rising action
• Climax
• Falling action
• Resolution

ChatGPT sees these components as crucial parts of the plot equation. It never deviates from this formula unless specifically directed otherwise. Therefore, the quality and originality of the plot will depend on the quality and originality of the data it has been trained on and the input prompts given to it.

The beings involved in the story. These can be people, animals, or other kinds of beings. Each should have their own distinct personality, motivations, and conflicts. By having a dynamic cast of characters, a story becomes more entertaining and relatable.

To create characters, ChatGPT can analyze the prompts given to it and produce the main characters based on three main factors:

• Character descriptions
• Relationships
• Goals

Once more, this AI model follows a specific formula for creating characters (albeit a simpler one than when generating a plotline). Do you see a pattern yet?

The time and place where the story occurs. It includes the physical environment, culture, historical context, and many more worldly components. The religions, myths, and legends of a setting’s sapient races all fall into this category as well.

When coming up with a setting, ChatGPT uses the following components:

• Time period
• Location
• Culture
• History

Due to a setting being much more physically defined than other essential elements of fiction writing, the AI model has an easier time coming up with tangible parts of the described world, such as noteworthy locations and the appearance of the landscape.

The point of view from which the story is told. It can be first-person, second-person, or third-person. It’s important not to swap between any of the aforementioned three unless you have a great reason. If you do, you may end up with confusing text that leaves the reader lost and frustrated.

ChatGPT tries to pick the perspective that it thinks is best for the prompt. However, this can backfire if you leave this element unspecified, which can result in ChatGPT mixing up perspective types. This ends in a narrative that’s all over the place and confusing.

The reason why the AI model has such a hard time with perspective is that it cannot quantify perspective like it can with other fiction elements. It has no formula for generating a point of view.

The central messages or ideas of the story. These can be straightforward and easy to grasp, or obscure and shrouded in mystery. Depending on your goals, either explicit or implicit themes can make your story more interesting—sometimes even both in a unique mix.

ChatGPT will be broad with themes, as it judges the importance of potential ones based on the prompt given. This ties into the components it analyzes from existing works. For example, if you get it to generate a story about aliens, it will use humanity’s innate fear of the unknown as a key theme.

Due to themes being story-specific, the AI model doesn’t have a true formula for generating them; it picks themes synonymous with those found in known works of fiction. It basically just copies the themes from similar stories unless you specify otherwise.

The spoken or written words between characters. This is the avenue through which characters reveal their personalities, relationships, and conflicts. Not all stories necessarily need dialogue, but it’s a great way to naturally incorporate important aspects of the characters. Furthermore, dialogue breaks up the prose in ways that result in better readability for most audiences.

ChatGPT uses traits of the characters in a story to generate dialogue between them. It also tries to keep a story’s context in mind. Here are the factors it will consider when creating dialogue text:

• Internal character traits (personalities, thoughts, and emotions)
• External character traits (speech patterns, dialects, and colloquialisms)
• Scene context

The AI model will try to advance the plot through dialogue as well as reveal character motivations. Keep in mind that it can be a bit on the nose and clumsy with such story components, so you’ll need to insert some subtlety via your prompt.

The author’s writing style. This includes word choice, tone, and the use of literary devices. It contributes to the overall mood and atmosphere of the story. An author’s life, upbringing, moral values, and more contribute to their writing style. However, don’t expect ChatGPT to perfectly emulate a famous writer.

Since an AI model can only work with what it’s given via details in your prompts, you need to refine them by making sure you speak ChatGPT’s language.

You can provide context in your prompts by including details about the setting; what the characters are seeing, hearing, smelling, tasting, and touching. Mentioning what your characters should be physically experiencing lets ChatGPT learn what the world is like without you literally mentioning everything in a location.

Incorporating flashbacks can be a solid way to reveal a backstory in a direct manner. When you tell ChatGPT to do this in a prompt, it will provide more depth and understanding of your story.

There are sometimes important story components that you can’t bring up directly. As such, using a bit of exposition can close unnecessary gaps in your world-building. However, you want to use exposition sparingly, as too much can result in boring text. Give ChatGPT a parameter for how much exposition to include so that it doesn’t get carried away. For example, include “Use two sentences of exposition” in your prompt for a scene.

Start by giving clear goals for a plot’s concept. Mention the genre, the main characters, and the central conflict or problem so that ChatGPT understands what you’re trying to achieve with your story.

Outline the major plot points in the story. Specifically, include the inciting incident, the rising action, the climax, and the resolution in your prompt. This is pretty much the exact formula that ChatGPT is looking for in order to answer your plot equation.

Provide clear character goals and motivations. ChatGPT needs to understand the goals and motivations of the characters in the story to build a solid plot structure. Otherwise, the story may come out feeling like it lacks a direction or that the plot isn’t progressing as it should.

Consider breaking down your story into three, four, or five acts. Segmenting the story and then giving it to ChatGPT via a prompt can keep things organized and balanced in terms of pacing and structure. You don’t want a story with an elongated beginning and short end—or similar imbalances.

Give ChatGPT detailed descriptions of your characters, including their physical appearance, personality traits, and personal backstory. This will help the AI model generate more nuanced and complex characters.

While mentioning a character’s goals and motivations is crucial for the plot, it’s equally important for developing the character themself. Guiding ChatGPT’s output in this regard will help your characters become more relatable.

You can also use individual character prompts to ask ChatGPT to generate a character’s backstory, a monologue from a character’s perspective, or a conversation between two characters.

Each character should have a distinct voice, with unique speech patterns, vocabulary, and tone. Provide ChatGPT with clear character descriptions and sample dialogue to guide its output.

Avoid using overly formal language. Instead, focus on creating dialogue that sounds like something real people would say. Try to emulate a conversation you’ve had recently to practice coming up with good dialogue prompts.

Dialogue should be grounded in the context of the story. Provide ChatGPT with clear descriptions of the setting, the characters’ relationships, and the current situation to help it generate dialogue that is appropriate for the scene.

Furthermore, dialogue is often more engaging when there is an underlying subtext. This can be achieved through the use of indirect language, sarcasm, or implication. People also lie a lot, so throw some of those into your character conversations as well for added realism.

In addition to getting ChatGPT to generate more interesting text, you’ll want to improve the style of such text. This is different than coming up with an initial style for ChatGPT with which to write fiction.

We briefly went over how to do so, but now we’re going to go into more specific details on how to accomplish style improvements.

Give ChatGPT detailed descriptions of the setting, characters, and objects in the story. Mention physical aspects of the landscape, the appearance of characters, and why particular objects are important.

Figurative language refers to metaphors, similes, and allusions. Such writing techniques are more subtle ways of getting across the importance of subjects, themes, and more.

Despite not being quantitative in nature, ChatGPT has a solid understanding of these techniques due to there being plenty of examples of them in existing works. Remember that this AI model learns from what humans tell it as well as what it has learned from available information about fiction writing.

Encourage ChatGPT to include sensory details, such as sights, smells, sounds, and textures, to help bring the scene to life. People are sensory beings; we perceive the world around us through what our senses tell us about our surroundings. As such, readers will become more immersed in the story if you tell ChatGPT to include such details.

You want to have varying sentence structures for multiple reasons. Firstly, it makes the text more interesting to readers. On top of that, the length of a sentence can help emphasize points or explain concepts in greater depth.

Shorter sentences—including sentence fragments—can make a simple point resonate in a reader’s mind. This can be done through both dialogue and prose. Conversely, longer and more complex sentences let you explore a point in better detail and allow readers to gain a better understanding of what you’re trying to say.

The first step in machine learning is to collect and organize data. In regards to fiction writing, ChatGPT looks at the vast volumes of existing written works—as we’ve mentioned a few times before in this article.

We’re stressing this point because you need to always remember that ChatGPT is an artificial intelligence that cannot do things on its own. Keeping this in mind will prevent you from getting frustrated when ChatGPT doesn’t give you the results you want.

Once the data is collected, it must be cleaned and preprocessed to ensure that it is in a usable format. This might involve removing irrelevant or duplicate data points, standardizing the data, and more.

After the data is preprocessed, it can be used to train a machine learning algorithm. During the training phase, the algorithm analyzes the data to identify patterns and relationships. This is done by adjusting the algorithm’s parameters until it can accurately predict or classify data based on the patterns it has identified. With that said, due to the amount of misinformation out there, ChatGPT is prone to sometimes generating inaccurate information—so always be wary and do your research.

Following the training process, the AI model must be tested to ensure that it can accurately predict or classify new data. This is typically done by using original data to evaluate the algorithm’s performance.

Lastly, once the algorithm has been trained and tested, it can be deployed to make predictions or decisions based on new data. When getting ChatGPT to write fiction, this last part comes about when you give it prompts.

As the AI model receives feedback and is exposed to more examples of good writing, it can learn to refine its output and generate more engaging and compelling stories.

Make sure you regularly give ChatGPT feedback so that it has a chance to learn what you want from it. If you don’t, you’re unlikely to get the results you want. ChatGPT can only improve if you help it. It cannot make improvements on its own.

You can fine-tune ChatGPT’s writing techniques by training it on a specific task. Get it to generate stories from a specific genre, like high fantasy or cosmic horror, and then give it feedback about what you liked and didn’t like regarding what it generated.

For example, if the high fantasy text it created didn’t have enough magic, give the AI model feedback like “include more magic spells” or “put more wizards and witches in the story.”

This can be done with all aspects of a story; not just the genre-related points. Use this method for the choice of words, theme implementation, character personalities, and more.

Here’s a summary of the main points we explored in this article:

• Remember ChatGPT’s capabilities and limitations
• Understand how ChatGPT uses fiction writing elements to generate text (plot, setting, etc.)
• Improve ChatGPT’s content by giving it solid prompts to work off
• Enhance ChatGPT’s writing style by giving it specific feedback
• Use machine learning techniques to help you get what you want out of the AI model

The old adage “practice makes perfect” is as true here as it always is. You’ll need to experiment a lot to get a suitable product from ChatGPT. It’s always learning and improving—therefore you should as well—not just by coming up with good prompts, but also by deepening your understanding of fiction writing. Keep reading stories from other authors and practice writing without ChatGPT.

The most important takeaway here is that AI models are a tool to help you improve; you shouldn’t expect them to do all the work for you.