ChatGPT “helped” me write a novel – and created a ruthless serial killer

What started as a Roald Dahl-style tale of the unexpected turned into what ChatGPT describes as a “psychological thriller” featuring a nighttime chase through a forest, writes Simon Jones.

The unexpected twist was that one of the side characters turned into a cold-blooded killer. That certainly wasn’t in the story arc.

Nor were the warnings from ChatGPT that the developing narrative had violated its own usage policies, due to the graphic violence and cold-blooded nature of the killings – with moral ambiguity, too.

How quickly ChatGPT drifted to the dark side!

I’d prepped the solid foundations for a fast-paced novel, and the plot and character consistency checks worked well. Running these checks after each chapter provided the opportunity to go back and correct minor inconsistencies – and to direct the storyline.

But then, somehow, we ended up on a huge ramble. The heroine was being chased through the woods in darkness – until then, the entire plot had played out in the middle of a large city, so that was a bit odd.

I let ChatGPT carry on and ended up with another 11 chapters, each around 2500 words, on a seemingly endless game of hide-and-seek through the woods. The TL; DR is that the hunters eventually become the prey, as a bit-part character transforms from PA to serial killer.

The novel began as a thought-provoking tale where you were supposed to develop a love-hate relationship with the main character. Somehow, along the way, it transformed into a chilling tale of cold-blooded murder in the dark.

It was created using ChatGPT 4o, so my next move was to ask o1-preview for a critique. This highlighted abrupt character transformations and a lack of consistency in themes and tone. Yes, that had become clear.

This first attempt won’t be riding high in the bestseller lists anytime soon, but there’s a tale to tell. It’s just a pity that ChatGPT isn’t advanced enough to help me tell it. Yet.

I asked ChatGPT o1-preview for its side of the story:

From my perspective as an AI language model designed to assist with creative writing, the evolution of the story into a psychological thriller with unexpected twists was a result of interpreting and expanding upon the prompts given. When generating narratives, I aim to introduce engaging elements, such as suspense and character development, to keep the reader interested.

In this instance, the introduction of a side character transforming into a cold-blooded killer and the shift from a city setting to a nighttime chase through a forest were attempts to heighten tension and explore deeper psychological themes. These elements can sometimes lead the story in new and unanticipated directions, especially without frequent guidance or specific constraints.

The inconsistencies and abrupt changes in theme and tone likely arose from the challenges of maintaining a cohesive narrative over multiple chapters. Without detailed outlines or continual input, the story can drift from its original arc as new ideas are introduced to keep the plot dynamic.

That’s encouraging. My guidance was frequent. More constraints needed on version two… which is probably a good rule for doing anything with a GenAI.