Editorial note: This article was originally published in the Knutpunkt 2026 book Seasons of Larp. It has been reprinted from there with the editors’ and authors’ permission. It has not been edited by Nordiclarp.org.
In 2022, ChatGPT was released to the world, and it immediately became the new hot thing. Especially in the beginning, before healthy criticisms were raised about where the training data came from. As I recall, it was especially when image-generating large language models (LLMs) made the rounds that it became clear that they were trained on actual artists’ work without consent or compensation.
This article is not about the history of LLMs or whether or not they are ethical to use (spoiler alert: they are not). Rather, it is a serious, cynical, and pragmatic look at the pitfalls of using them. I realise that many people have started using AI-enhanced tools in their work, and it seems, as The Matrix’s Agent Smith said, “inevitable”.
I felt inspired to write about this for the Knutepunkt book already in 2023, for the call for articles for the 2024 book. Organisers had already started using ChatGPT to create characters at larps, and even in my own projects, co-organisers used LLMs. However, because of the justified backlash, I withdrew my article. I did not feel comfortable tackling the topic.
To be honest, I still do not feel comfortable. I am not sure anyone is capable of discussing this without inevitably ending up looking back and looking like a complete idiot five years from now.
I am now going to be that idiot and put some thoughts on the table, with every intention of being disregarded or proven wrong.
Why use ChatGPT?
Let’s start out with some comparisons. When I started creating my own characters for Dungeons & Dragons decades ago, I struggled to come up with good names. A friend showed me something: a fantasy name generator. It was a game changer, a fun tool to get over the parts I struggled with. I could just press a button and it would reroll new names until I found something I was happy with. So I get the appeal with ChatGPT.
Frankly, it was a relief and in a way it got me more excited about working with characters and finding similar tools to support this path. When I started writing my first characters for others, I also used different support tools to create them. An obvious tool is some sort of autocorrect. I struggle with spelling, both in my native Danish and in English, so having word processors help me out with spellchecks and give grammar corrections is a big help. Having Wikipedia and Google to help with historic contexts and explaining words I don’t understand is absolutely amazing.
ChatGPT and similar LLMs are in a sense the newest kids on the block, an amalgamation of these tools and information databases. I get why someone would get excited by this new tool and try to integrate it into their work and personal life to become more efficient.
So let‘s put it to the test! I want to discuss and explore the use of tools like ChatGPT in larp design from the lens of character writing.
Surrendering personal growth
I had a classmate who, when he was writing in Word and autocorrect made a suggestion to correct a mistake, would delete the entire word or even sentence and write it from scratch. At the time I felt this was silly; you could just click once to replace the word with the one word recommended. What I realized many years later was that this was actually intentional. He used the autocorrect to spot the mistake and then correct it, building muscle memory for the correct spelling of the word and taking the slower path to actually improve.
All these tools that we have for helping us out work the same way. They give you a seemingly correct answer on the fly and it is easy to just accept it and move on to the next task. If you develop a habit of using these tools on autopilot, you hamper your own development and also miss the mistakes the tools make, maybe due to your own incorrect input. No matter how advanced and smart LLM or AGI or whatever comes next is, you still need to use the tool correctly to get the correct output.
Put simply, if I input 2+2 in a calculator, it will answer “4”, even if I meant or intended to input 3+3, which I will now assume is 4 going forward. In a small and simple question like this I might be able to find the flaw myself before moving on, but the more complex the tool, the harder it is to easily spot the incorrect input.
Raising a question and arriving at an answer that you actually understand is an opportunity to be a fruitful journey with unexpected personal growth, if you let it. Seeing how my classmate chose to learn from a tool instead of letting it do all the thinking taught me something.
Arriving at actual character creation with ChatGPT
I have not used ChatGPT for creating characters, but I have used something else: actual people. In a few projects, I have been lead character designer for larps with over 100 characters; on that scale it is difficult to do all on your own, especially if you want to give each character your full attention. What we often do in the bigger larps is have teams of character writers create the characters.
People, just like ChatGPT, need directions on what to do and how. If you want clear and meaningful character, you need to be able to convey to them the framework and assign clearly defined tasks. Similar to using a prompt for ChatGPT, you need a design guide for your character writers. Now, real people take much more time than ChatGPT, not necessarily because they are slower, but because they don’t have as much free time as ChatGPT. I have seen people in the zone write brilliant stuff as fast as ChatGPT can spit out a text. However, co-writers are not available 24/7 and they are usually not motivated to do as many rewrites and complete changes in direction as a mindless robot is.
Accountability
But people have something really, really cool: accountability. If there is a problem with the character or something you don’t understand, you can talk to the actual writer about what they actually mean to a degree that a chatbot is not really capable of. Real people can ask you questions about whether you understood the text they wrote or got the underlying ideas they tried to convey. This conversation can also be extended to the participant. If you go to the larp and someone has gotten a character written by your writer, when the player comes with questions you cannot meaningfully answer, you can direct them to the actual writer. This is one of those things that ChatGPT will likely never be able to provide. It can create a new analysis of its own text, but it will not be able to go back and go through its process or the intent behind it that might be lost to the reader. It can only provide a new answer and another new answer and so on, until you are satisfied or give up, with no guarantee that it ever gets close to your original task.
This, as I see it, is the true challenge of using ChatGPT. When you arrive at the larp and players come to you with questions about the character that was written for you by AI, you will struggle to help them with their character because you don’t really know any more than your players. Usually as a character designer, you have some bigger thoughts and ideas than what is conveyed in the text, and for some players that broader perspective can be the key to understanding how to actually play their character. This is also a problem that ChatGPT will likely run into: understanding and conveying the bigger picture of the larp and how the written character is supposed to be a guide on how to unlock and interact with your larp. ChatGPT will just write you a cool story, but getting it to create connections with other players and leave open options for creating plot and more connections is something it cannot do. It will leave you as an organiser and designer without answers for your players.
Going into the ChatGPT generation
With that in mind, now it is time to test it out for real. What would ChatGPT do if I put it to the test for creating a character and how would I rate it and the process of working with it? I have no idea what will really come out of this, but I have a load of assumptions we are going to explore. Therefore, I also want to keep it fairly open-ended and generic to not direct it too much.
This is the prompt I put into ChatGPT:
“I would like you to help me create a character for a magic school larp driven with social conflicts and the stress of school life. Modern Urban Fantasy world, close to our own world. The character needs to also have relations with other people in the larp.”
Chat GPT then asked some follow-up questions about style of play and, funnily enough, also asked about the name and pronouns. I wanted it to focus on narrative and keep it story-driven and so, I said it could choose name and gender itself.
This was the result:
Character Name: Iris Vale
Age: 17
Pronouns: She/they
Magical Speciality:
Resonance Magic – a subtle and emotion-ally charged art that lets Iris sense and manipulate the emotional “frequencies” around her. She can make someone feel slightly calmer in a crisis or amplify the tension in a room until it feels unbearable. It’s powerful in theory — but dangerous for someone as emotionally raw as Iris. Strong feelings can rebound, causing unpredictable outbursts or draining exhaustion.
Background:
Iris Vale grew up on the edge of the magical world — her mother a once-promising enchantress who lost her license after an “incident,” and her father a non-magical journalist who still writes angry columns about “elite magical institutions.”
When Iris received her acceptance letter to Orchard Hill Academy of Arcane Arts, it felt like a second chance — not just for her, but for her family name. She entered determined to be better, quieter, safer than her mother ever was. But magic doesn’t behave when you try to suppress it, and neither do emotions.
At Orchard Hill, Iris found herself trapped between worlds. The pureblood students treat her like charity. The non-magical-borns think she’s cold and privileged. She spends long nights in the Resonance practice rooms, whispering apologies to the echoes that only she can hear.
Personality:
Soft-spoken, observant, and almost pain-fully empathetic, Iris feels too much and shows too little. Her fear of losing control makes her withdraw just when she most wants to connect. She keeps meticulous notebooks about emotions she can’t express aloud — cataloging smiles, stifled sighs, every shift in tone.
But when she does open up, it’s like a dam breaking: poetic, intense, and impossible to ignore.
Motivations and Inner Conflict:
- To belong — Iris desperately wants to feel accepted, but fears she doesn’t deserve to be.
- To control her emotions — Her magic is strongest when she feels deeply, but that’s also when it’s most unstable.
- To redeem her family’s reputation — She hides her mother’s scandal, but it shadows everything she does.
- To understand “truth” in emotion — Part of her magic is listening, but she’s beginning to question whether it’s ever ethical to hear what others don’t choose to share.
Secrets:
- Iris’s mother didn’t just lose control of her magic — she caused an incident that led to another student’s death. The school quietly covered it up.
- Iris sometimes feels the emotions of the dead student echoing in her resonance spells. She doesn’t know if it’s guilt, trauma, or a haunting.
- She’s started experimenting with using her magic to suppress her own emotions — a dangerous and unsustainable co-ping mechanism.
Connections / Hooks:
These are open-ended, but with enough flavour that other players can easily attach themselves:
- The Confidant — Someone who sees through Iris’s polite mask. Maybe they share a quiet bench by the greenhouse after curfew, talking about things they shouldn’t. Iris trusts them more than she should.
- The Rival — A pureblood student who knows about her mother’s scandal. They weaponize subtle gossip, but perhaps it hides a strange fascination or even attraction.
- The Anchor — A non-magical-born classmate whose emotional “frequency” Iris can’t quite sense — which makes them peaceful and terrifying at once. They remind her what it’s like to just be.
- The Mirror — Another student struggling with control — but who embraces chaos where Iris fears it. They tempt her to stop suppressing her emotions and to feel everything.
Roleplay Prompts:
- You often feel the emotions of others before they even speak. How do you act when someone’s sadness seeps into you during class?
- Someone discovers that your magic can manipulate feelings. How do you convince them you’re not dangerous?
- Late one night, you hear the same emotional “frequency” your mother used to carry. What do you do?
- You realise one of your classmates is using you for information or influence. Do you retaliate, withdraw, or empathize?
Tone:
A character made for intense, emotional, human play. Great for moments of quiet vulnerability, unspoken attraction, and moral questions about empathy and consent. Iris isn’t about power struggles — she’s about feeling too much in a world that punishes that.
Feedback on ChatGPT character
This is what ChatGPT created in the first try after a bit of clarification. It also offered doing more for me, but I think we already have a lot to work with.
Looking at the full character, I have to be fair: I didn’t give it much coaching about what kind of larp I was creating and some of the clear problems could have been avoided by prepping it better. However, I wanted to see what it would create from the ground up instead of showing it a finished character it could copy.
Now, I will go through each section:
Resonance magic: It is really creative, but as it is described, I think it is going to be difficult to actually do in play at a larp without complicated game mechanics (which I told it to avoid) or the larp being very internal-immersion. It is not impossible to play with, just a challenge. I can also already see a player coming back to me with a lot of questions on how I see this actually working in the larp and how to play it out.
Background: Very cliché, with outsider parents, her representing a second chance, and conflict with pureblood students. Lots of good stuff. So far also very difficult though, without much support from anyone. As I see it, this is going to be a very lonely player.
Personality: This is already a bit better, because it has a mix of things. While very introverted and likely very relatable for a lot of inverted magical fantasy nerds, the character still has some concrete things to do, like write a notebook and have dam-breaking moments of intense emotions and poetry. The notebook is a great prop and notes can be forgotten or stolen to give play to others. To have an excuse to prepare these intense scenes of finally opening up to someone is something I think could be fun to do in a larp. Writing too introverted characters is always dangerous because they easily just become wallpaper for other players’ stories. It is very important to give them ways to avoid that.
Motivations and inner conflict: This part is very bland and already made obvious from the rest of the character. I think with a longer character, this part would make sense as a summary. Now it looks like something that was just added to take up space.
Secrets: This part adds the fact that her mother’s incident caused the death of another student and it was covered up. This is a big secret and could create an interesting story. It could be a major plot point for a larp, since the school covered it up and so other characters need to be involved with it. This is also one of those things that you as a larp designer could easily miss after just reading the first half of the character and forget to include elsewhere, and then this player would be looking for a mystery that only they know about.
I like the other secrets, about using her spells to connect with a dead student and also experimenting with using magic to suppress her own emotions. I am not sure they fit as a secret and they should maybe have been part of her inner conflict. Both ideas could lead to interesting play.
Connections / hooks: I asked ChatGPT to also create relations with other characters, and it did ask me if I had those characters or if it should just create some generic hooks, which it clearly then did. It even kind of explains that it is work in progress: “These are open-ended, but with enough flavour that other players can easily attach themselves”. This is a place to start and obviously more work would need to be done. Which once other characters had been created it would be easier to ask it to link them together.
Roleplay prompts: So, now we humans are also getting prompted by the AI on how to play. I like it. I think they are actually fine. There is some stuff to try to make happen at the larp and they are all ideas to explore with co-players in a workshop.
Tone: “Iris isn’t about power struggles — she’s about feeling too much in a world that punishes that.” What a banger ending! I think this works great at getting someone excited about playing this character.
This is again clearly the strong suit of ChatGPT, it is great at glazing and making things sound interesting. However, it lacks consistency and understanding of how things would actually work out in our physical world.
My conclusion
Using ChatGPT is fun, no doubt about it. It feels great just giving it a prompt and seeing it get to work without batting an eye. The process of not getting any pushback or questions is very different from how you work with co-creators that are humans. ChatGPT will provide no pushback, no second opinion, and no time limits.
Let me say that again: “No pushback, no second opinions, and no time limits.”
Depending on what you get out of the line above, it will say a lot on how you view and work with other people. I can say from personal experience, that working on creating the first College of Wizardry, there was a lot of pushback, tons of opinions, and never enough time. It was frustrating, it was tough, and it felt like the most impossible thing to put together. There was external feedback, so much arguing back and forth internally, and every part of the design was discussed endlessly.
These discussions and arguments, as draining as they could be, formed the core that held the larp together. Having real discussions with real humans is how we build a proper foundation as we poke at problems from every direction. They prepare us for when the larp meets the reality of actual players standing on location and asking questions about how to actually play the larp.
In essence, I think my conclusion is that even if ChatGPT can do some of the work, you will still need to take the time to go through and process what it has cooked up. Ideally, you discuss the content in a group with diverse perspectives and make sure it works for your larp. In the end, it is not really a great path to avoid putting in the work. ChatGPT will only be able to provide you with a starting point.
If we choose to use ChatGPT for character creation, we cannot remove the human element in checking and editing. In the end, we will just be glorified editors for ChatGPT’s attempts at creativity. Maybe it is fine to just skip the middleman and go directly to the creative part ourselves.
So here in the end, do I feel more comfortable with discussing this topic?
The short answer is “Hell no!” There are a lot of legal and ethical issues that I did not want to delve into in this article. I feel it is already way too long. And it is easy to write it off on one side of the fence or the other.
But I have now written and I am now that idiot with some thoughts on the table, with every intention of being disregarded and proven wrong five years, or maybe already five months, from now. And I now pass the hot potato to you:
What are your thoughts on using ChatGPT in larp? Do you use LLMs in other situa-tions?
This article is republished from the Knutpunkt 2026 book. Please cite it as:
Nielsen, Charles Bo. 2026. ‘Using ChatGPT in Character Creation.’ In Seasons of Larp eds. Erlandsson, A., Svanevik, M., Holkar, M., and Brind, S. Gothenburg. Knutpunkt.
