AI/LLMs Murder Hobos are out, Paladins and Bards are in, according to the data

@Wold Newton Thanks for the information!

You mention 54.7 million words, but I have no feel for how many campaigns that actually is, and I do worry about having a good representative sample. (Or at least as good as one as possible as the self-selection bias amongst those who use the tool as AI is somewhat polarizing among RPG hobbyists.)

Can you let us know how many separate campaigns that is? Perhaps of some minimum length (at least 3 sessions?)

Thanks!
 

log in or register to remove this ad

1) Murderhoboism has nothing to do with the alignment on the character sheet.

2) Alignment as an indication of behavior is very poorly corelated. I find in general that people justify all kinds of terrible behavior anyway.
I do find that what is on the paper does not relate always to how the character is played. I have known players to always play a certain way or style and have all sorts of classes, skills, backgrounds, alignments, and such.
 

Tech Bros Release Gross Speculation Based Off Minimal Data; Claim "Extrapolation Generates Engagement."

New analysis of small pockets of data show that you can make ridiculous claims on the internet, but saying it's based on AI is the real ticket to creating traction and getting free advertising.

SAN LUIS OBISPO, CA, August 24, 2026 – Archivist AI, an AI scraping tool that claims to have absolutely no knowledge of D&D Beyond, Roll20, or screen-free play, has discovered that they can cheaply advertise their platform by making absurd extrapolations based on minimal data.

“We didn’t set out looking for this – it just kept showing up in the meetings,” said Greg Wood, co-founder of Archivist. “The idea of generating real content to promote our product was simply overwhelming. Have you ever tried to produce real data analysis that people care about? It's exhausting. And the cost of generating charts and graphs is just too much. We considered getting into the infographic game, but decided it was too saturated." When asked how he ended up with an em dash in his quote, Greg declined to comment.
 


Some interesting numbers there.

1) A slow creep up by 5.5, although base 5e still outnumbers it 50%-21% (as of Q3 2026). I imagine quite a few of "base 5e" tables are like mine and use a mix of both, though. (I imagine the system in use is self reported, and people using both just pick "5e".)

2) Cleric as the number 1 played class at 11%. Not an obvious choice, but considering that cleric is still the assumed healer, and most parties will want some healing, not a huge surprise. The sheer variety of subclasses I would assume also plays a role.

3) Warlock being 12th is also a bit of a surprise to me.
 

LOL. To be fair though—I've been using em dashes for a long time. I'm kind of miffed that it's now considered a tell of AI generated text. Maybe AI has spent too much time scraping my blog.

Sure, you write with em dashes. But do you speak with em dashes? Is there an accent that sounds more like an em dash than a period, a comma, or a semicolon?

What's more likely: Greg made some sort of overly dramatic pause in his quote where the author knew it had to be written as an em dash, or the quotation marks are complete BS and the entire interview is AI slop?
 

It would still just make less good works that use too much resources and serve no real purpose while making it harder for creative humans to make a living.

I don't want GenAI goods, I want GenAI services I can use without feeling icky.

E.G. I spend a lot of time on game notes. As a tool for transcribing notes, turning chat threads into readable documents, and other summarization tasks it would be great.

Also, as someone who sometimes has issues making up distinct NPCs when my players make a hard left, I'd like to use gen AI for "generate 3 notable NPCs for the town of Blargawizzle in the region of Hackenfort" in lieue of dice-based NPC generators, especially if it were smart enough to know the demographics of Hackenfort.

So I want a genAI secretary, but without an ethically sourced LLM I won't have a genAI secretary.
 


I don't want GenAI goods, I want GenAI services I can use without feeling icky.

E.G. I spend a lot of time on game notes. As a tool for transcribing notes, turning chat threads into readable documents, and other summarization tasks it would be great.

Also, as someone who sometimes has issues making up distinct NPCs when my players make a hard left, I'd like to use gen AI for "generate 3 notable NPCs for the town of Blargawizzle in the region of Hackenfort" in lieue of dice-based NPC generators, especially if it were smart enough to know the demographics of Hackenfort.

So I want a genAI secretary, but without an ethically sourced LLM I won't have a genAI secretary.
Those functions exist without LLMs, though. Machine learning comes in different forms, and more "limited" ai can do a lot of that sort of thing by being built from the ground up to do a specific type of task.

If we have a computing hardware breakthrough and LLMs suddenly can run efficiently without generating massive waste heat and using nations worth of power, yeah, ethical uses could arise. But the ecological and economic cost makes it impossible right now to have ethical LLMs at all.
 

is their sample size big enough to be indicative of anything other than its own users?

I do suspect the theatre club generation has embraced the narrative storytelling style of play v the old skirmish level wargame and that 5e rules certainly cater to it. Great to see the noble Paladin getting some love but I'm surprised alignment is being highlighted as indicative of anything
 

Remove ads

Top