You can use AI in multiple ways to help you as a writer. The most foundational one would be to use it as a partner to bounce ideas off of, discuss specific parts of your story, or as an analyst who checks for inconsistencies and plot holes. Brainstorming with AI as an author can help you come up with lots of new and exciting story elements.

Here’s how AI can help with brainstorming different parts of a novel:

Character Development: AI can be a big help when crafting complex characters. With AI, you can ensure the dynamic development and evolution of diverse character arcs, by looking at their goals and needs, and how they attempt to attain them. If a part of the story doesn’t make sense, because a character acts in a way that doesn’t make sense for it, the AI will know.

Plot Twists and Conflict: AI is great for creating unexpected plot twists that even surprise the author, keeping readers engaged. It can also make the conflicts in the novel more genuine and reflective of the themes and genre. It’s also much easier to make sure that scenes have the appropriate amount of action and tension for the pacing of the story beats you try to hit.

World-Building: AI can assist in building immersive worlds for storytelling, from futuristic sci-fi cities to enchanted magical forests. AI can also help design complex societal structures with specific political, cultural, and historical attributes. You can work on conlangs with it, or flesh out existing world-building by filling it with more details, which in turn can lead to new ideas for your story.

Dialogue Generation: Dialogue brings character voices to life, and AI can create authentic conversations that capture their uniqueness. It can experiment with different language styles to make sure the dialogue matches the setting and time period of the novel. Although, granted, this is still in the beginning stages, so you need to do a lot more aftercare for dialogue than for narration.

Theme Exploration: Themes give a story its heart and soul, and AI can help find and blend different themes into the plot, checking how they impact the characters and story to make sure everything fits together seamlessly. It can also help you maintain a consistent theme throughout the story.

Title and Synopsis Creation: You need a captivating title and an engaging synopsis to hook readers and publishers. AI can play a crucial role in coming up with titles and summaries that capture the heart of the novel and grab people’s interest, because, let’s face it: titles, blurb, loglines etc require sales skills, which we might not necessarily possess as writers. I know I don’t.

Subplot Development: AI can help make the story more impactful by adding detailed subplots that can explore the setting, juxtapose the main character’s story, and enhance the pacing of the novel.

Research Assistance: Research makes a novel believable. AI can make research easier by giving you info on history, science, or culture to make sure your story is authentic. It can also check the text to see if there are mistakes you’ll have to fix, and it can make suggestions to better implement certain facts.

Revision and Editing Suggestions: AI can give helpful feedback on how to improve your writing, like suggesting better flow, structure, and language, which can make your manuscript even better. This is more relevant for newer writers.

So, brainstorming with AI can speed up the idea generation process and introduce authors to a broader range of unexpected ideas, resulting in more creative storytelling.

Using AI in the writing process is like going on an adventure that can really improve your brainstorming.

Here’s how to get the best creative results from this partnership:

Start Broad, Then Narrow Down: Start with big ideas to get the creative juices flowing, then gradually narrow down to the best ideas from the AI.

Be Specific in Your Prompts: Paying attention to the details is key. The AI will give more accurate responses if you provide a clear prompt, but you might need to give more details to get the results you’re looking for. This part of the process is trial-and-error. Finding out how much the AI needs to know, and how much is too much, can take a while, depending on the case.

Use Iteration to Explore Variations: Take advantage of the AI’s ability to quickly tweak creative things. Just a few minor adjustments to the characters or story can lead to so many creative options. If a character has potential use for more conflict, the AI will know. If characters can be merged to tighten the story flow, the AI can help with that, too.

Feedback Loop: Talk to the AI and give it clear feedback on what it suggests. This feedback loop helps us keep improving our ideas. Think of it more like a feedback partner than a vending machine. Accepting the first idea that comes out is never a good strategy.

Combining Ideas: Fusion is your friend. Combine AI-generated snippets in unexpected ways for some fresh and unique concepts. We already do this when we come up with book ideas. Jim Butcher, for example, once threw together “Roman Empire” and “Pokémon” to create his Codex Alera series. With AI, you can rapidly preview results of such fusions and see what does and doesn’t work before you waste time developing it into a story.

Ask for Alternatives: Encourage the AI to provide alternative ideas. Don’t be afraid to try different prompts to make your work more interesting. You can also experiment with AI temperature settings (for example, in OpenAI’s Playground). Just generate a bunch of ideas and bash them together.

Balance Detail and Creativity: Balance out detailed instructions with some creative freedom. Don’t be too specific, let the AI’s imagination run wild. As mentioned above, finding the right mix of specificity and creative freedom can take a while.

Incorporate Randomness: Add some randomness to boost your creativity. You can do this by using random word prompts or asking the AI for creative ideas that loosely relate to your theme. “Give me 5 ideas on how this could play out” is also a good way to get some variety and more interesting plot twists.

Consistent Checkpoints: Check the AI-generated content regularly to make sure it fits with your novel’s vision and make changes as necessary. The farther you get, the higher the error rate, which has to do with AI’s memory limitations, which we will discuss more in-depth later.

Human Intuition: Humans are irreplaceable. Let your instincts guide you in choosing which AI suggestions match your creative vision. Just because something makes sense doesn’t mean it also feels right. Plus, the AI could have missed a clue, or you didn’t make it obvious enough for it in the first place. AI is like a toddler with ADHD sometimes.

Set Clear Goals: Set specific objectives for each round of brainstorming to keep it on track. If you don’t, you will end up in Lala Land, with a huge pile of fun-sounding ideas but no clear concept. The goal needs to be concrete: Let’s develop this character. Let’s talk about how blocking would work out in this combat scene. Let’s look into the culture and history of this specific place. That’s how you go about brainstorming.

Know When to Step Back: Sometimes, you just need to step back to move forward. If you hit a creative block, give yourself some space. But not only you yourself might have run into a wall, AI is very much able to do that, too. In that case, start a new chat. AI sometimes goes down weird paths, and once it’s where it’s at, it can’t back down. A fresh chat session often fixes these issues.

By following these tips, authors can harness AI as a creative partner, perfecting ideas with a blend of science and artistry. Working with AI can make characters more interesting, plots more exciting, and stories more immersive.

When I talked about consistent checkpoints earlier, I mentioned certain limitations of AI. What I was talking about is the AI’s short-term memory, the context window. Knowing the context window is key to collaborating with AI in writing, and there are tools for this I will write about in a different blog post. Here, I’ll discuss what you can do within, say, ChatGPT or Claude’s chat.

Here’s how to navigate it:

Prompt Design: Crafting prompts that fit within the AI’s context window is critical. Include all the important info so the AI stays on track, but make sure not to give it more than it can keep in mind. Imagine a magnifying glass wandering over a page. That’s what the context window is. It can only see what the glass is currently hovering over, while the rest is going to blur until it’s eventually invisible.

Segmented Brainstorming: When brainstorming long stories, divide your ideas into smaller parts that match the AI’s context window. This helps generate relevant suggestions for each segment. It’s also important to not overwhelm it with irrelevant info. If you’re brainstorming a certain character, it doesn’t need to know every detail of certain locations or other characters. It has to know how things relate, though.

Consistency Checks: Make sure everything in the story makes sense, because AI might forget things from earlier, as discussed above. It’s a good idea to keep a separate document for important story elements. Character profiles, location documents, lore texts.

Focused Sessions: Just focus on one thing at a time – like the story, the setting, or the characters – to generate more detailed content in the AI’s context window. This harks back to the “Segmented Brainstorming” part. AI only has so much memory, you have to be selective and restrictive.

Iterative Refinement: Make the most of the AI’s context window to gradually enhance the story. Once you’ve come up with ideas for one part, use the feedback to guide the next, so it all flows smoothly. You can run the same prompt multiple times, and you can tell the AI to read its own output and iterate on it. Tell it to grade itself on how well it executed your prompt, and ask it to do it again and improve itself. You’ll be surprised how good AI can get if it gets the chance to re-do a task.

Writers can use AI to enhance their storytelling process by managing the context window strategically, making it more efficient and enjoyable.

You need to plan and document things carefully to work around the AI’s limitations.

External Documentation: Have an external “story bible” to keep tabs on all the important stuff like character traits, plot lines, and world details. Look at this document to remind the AI of what’s been happening in the story. See “Consistency Checks”. You’re probably doing this already, anyway, but it’s especially important when working with an AI that doesn’t even remember who you are the next time you talk with it.

Iterative Summarization: Have the AI sum up the brainstorming ideas into core summaries at the end of each part of the brainstorming session. Summaries are one of the key elements when working with AI. Not only can you use it to “externalise memories”, you can also use them to go through the same problem repeatedly until you get the desired outcome. This is extremely useful when working on structural problems, like an outline or scene beats.

Contextual Cues: Reference previous story elements in your prompts so the AI stays consistent. I already mentioned that there are tools that can already do this, but we’re still talking about the basics here.

Cycling Key Information: Make sure to occasionally bring up important info in your prompts so the AI doesn’t forget what’s going on. This is very important in ChatGPT or Claude, as these tools don’t have access to anything outside their context windows, whose size can be deceptive. As of November 2023, Claude can hold 100k tokens, while GPT has 128k. Calculate with a ration of 1 token being 70% of a word, which means in theory, Claude can remember 70.000 words. But that’s not how it really works. It’s most accurate at the start and end of the context window, while everything in the middle gets blurry. Because it’s so focussed at the beginning and the end, you will see better results by bringing facts up regularly, so it “pulls” them from memory.

Using Keywords and Tags: Come up with some keywords or tags to remind the AI of the story. AIs think in tokens, as mentioned earlier, and a longer name like “Mister Fluffypants The Third” eats up 8 of them. If you assign the tag “Fluffy”, you reduce that to 2. You also make it easier for the AI to reference things in its memory.

Integrate these strategies to make the AI work well with your evolving storyline.

There are now tools for authors with integrated AI, and I’ll introduce you to what I think is the best one currently, in my next blog post.