The 13th Quest

(This review was originally posted on the IntFic forums during ParserComp)

So, generative AI in IF, there’s a topic I’m sure we’re all super excited to get into! Three years on from when ChatGPT was unleashed upon the world – and the same day that Elon Musk’s Grok chatbot went on an antisemitic spree before they hastily pulled it offline – the only thing I find more enervating than the discourse is leaden LLM prose. And yet, here I am talking about it, because the ABOUT text for 13th Quest reveals that the author used ChatGPT “to improve some of the text descriptions and responses” – and those “improvements” mark a noticeable departure from the author’s previously workmanlike oeuvre, while also calling into question what exactly we want to get out of an old-school puzzler.

We can get through the non-LLM stuff pretty quickly, since 13th Quest is of a piece with the author’s other dozen games, at least from the ones I’ve played: it’s written in a robust custom system with one or two quirks that have long since become second nature (you can’t interact with things in containers until you take them, is the big one); there’s a big book you get at the beginning that intersperses some key puzzle clues between a bunch of lore; there’s a magic teleporter that whisks you between self-contained areas; and you have a host of medium dry-goods puzzles of consistently-solid quality that are clearly more the point of the exercise than the MacGuffin quest that provides the notional story.

Speaking of the narrative, this time out there’s a Celtic theme that emerges about a third of the way through – after somewhat-bewildering trip to Damascus – that provides some nice faerie flavor, as well as a well-implemented leprechaun NPC who wants you to perform a half-dozen fetch quests. This makes for an engaging middle of the game with relatively clear goals, and some neat set pieces as you venture through such classic fantasy settings as a brambly forest maze or a sweating-hot volcanic mine. On the down side, despite some stirring words in the ending gesturing towards the quest being more than a simple treasure hunt, the opening dumps you into the game with a shrug, saddling you with amnesia and a letter that mumbles its way through some empty verbiage about pacts and bloodlines and threshold without establishing what you’re meant to be doing and why it matters.

And here’s where the LLM question comes in, because, annoyingly, in 2025 I can’t read vague fantasy nonsense in a game that says it used ChatGPT without wondering whether that vagueness is an intentional choice, or at least an honest mistake of craft, or just a symptom of a bullshit machine trying to run out the clock. Thankfully, the game’s prose isn’t completely affected with AI-speak – as I mentioned, there’s some evocative fairy-tale stuff like the location in the ice-realm labelled as “the Hoarfane” – but every so often I’d come across a phrase that would set my teeth on edge twice over, once for being kinda bad and a second time because now I had to think about whether it was human-bad or not:

"The material [of the letter] is thin and brittle, edges both frayed and curling, like it’s waited years for unfamiliar hands."

"You see no windows but adequate light coming from an unknown source illuminates the location quite adequately."

"The silence is both heavy and profound."

Even the heavy grimoire, which is always a highlight in these games, seems saggier this time out; it tells one of its stories twice over, which could be a way of showing how fairy tales twist in the telling, or could be an LLM losing track of things. And it’s not just the prose – there are some undescribed exits and unimplemented bits scenery that could be an indication of AI use impacting the gameplay.

Again, I don’t like grumping out about this stuff – I’d much rather be spending time talking about the leprechaun, say, whose role in the plot is never explicated but who’s got surprisingly detailed things to say about just about any random inventory item you show him. But since IF is made of words, I tend to pay a lot of attention to them! And beyond that, nostalgia-bait treasure hunts are one of the IF subgenres least robust against the question “wait, is this just filler?” Of course, part of what’s fun about the old school is that a good puzzle can just be a good puzzle, without needing to be part of a clever unified gameplay system, or provide important a thematic counterbalance to the events of the narrative. But a risk of the style is that it all the combination-guessing and keycard-fiddling can feel arbitrary, just there because players expect a certain density of Extruded Adventuring Product before they collect their last plot coupon.

I don’t think 13th Quest does badly on this score, to be clear – there are lots of recognizably human touches, and despite a few overly-obscure clues or slight frustration with the parser, the puzzles are entertaining enough to work through. But this is a bad line of inquiry to put the player onto regardless, all the more so because from my viewpoint the addition of ChatGPT-authored prose feels like a net negative even on its own terms. The game would do fine standing entirely on its own feet, and I hope the 14th and 15th and 16th quests go back to the old way of doing things, so I can too, instead of having to come up with yet more ways of writing “god, I’m tired of writing about LLMs.”