Lindenbaum AI policy

 Hello gamebookers!

The times, they are a changin'. I mean, they are changing too fast for anyone to keep up with these days.

Now that Lindenbaum '26 is around the corner, the judges and I had a little think about AI. The catalyst for this was last year's winner, Pádraic Henrysson-MacOireachtaigh, who wrote An Island who wrote in his acknoweldgements that he used ChatGPT and Gemini to playtest his book.

Here is his quote about that in the acknowledgements section of the book:

"I used AI tools (ChatGPT and Gemini) as early ‘playtesters’ and occasionally to refine ideas – but the mechanics, story, and words are as much my own as a creative work ever is."

So the wave of AI has hit the Lindenbaum competition and we needed to think about how to deal with it.

My main aim is to make sure that AI generated books don't take part in the competition. Pádraic wrote the book himself and used AI to improve it.

First, I tried to see how good an AI gamebook could be.

I tried to ask Chat GPT to make a gamebook using the Lindenbaum rules as a prompt. Here is its output Lindenbaum 2026 Guide.

If you don't want to read the thread, the gamebook produced was very minimalistic. I tried to ask it for at least 80 sections and it couldn't manage more than 40. Its sections were also very terse.

Basically, if that book was entered, I think it would get no votes from both judges and readers.

So no danger of a pure AI gamebook winning. I still don't want people wading through AI slop entries though.

The trouble is that there is no 100% way of spotting an AI generating book and I don't want a false positive. However, if a gamebook does not function as a gamebook, then I feel that I can disqualify it from entry. What does it need to do to function as a gamebook? It needs section choices that you can turn to and those sections need to be intact. 

There are things that just make a good gamebook in general:

Systems where the probability of success and failure are both significant. If it's impossible, it's frustrating because it can't be played fairly. If it's impossible to fail, then it's frustrating because it's a waste of time.

Section links that work.

EDIT: Enough meaningful decisions - Gamebooks need meaningful decisions. This means that most sections should end in 2 or more options that both lead to a meaningful consequence. Of course, some sections will be death sections, some might just be descriptive prose to move between scenes and some might have only 1 option that doesn't lead to sudden death. These are all valid choices.

However, if a gamebook has too many sections that just lead to 1 other section, that does not give enough meaningful choices. Also, there are some gamebooks that were clearly written as novels first and then some options were added in later. However, since the "main storyline" has already been worked out, options that deviate from that main storyline either lead to sudden death or find some way to just loop back to the main storyline, which means that it is not a meaningful decision.

Proofreading errors that make the prose hard to understand.

So, if we're not going to stop this wave, I'm just going to have to surf it. For that reason, I'm not banning AI. I'm banning nonfunctioning gamebooks.

So, what can AI be used for, if not for writing?

Here are some suggestions that AI can be used for:

Looking for broken section transitions

To make sure that all of the rules are being used an used as intended

Making sure that the difficulty of the book matches the difficulty that you would like to communicate

Playtesting

Help with translating to English

The aim is that everyone can make their gamebooks the best they can be. The biggest tragedy is an entry that has lots of great ideas, but is let down by poor statistics or proofing or broken links. AI is able to clear up all of those problems and augment peoples' ideas and writing.

I look forward to all the entries in 2026!