AI in TTRPGs: Automate the Admin, Not the Story

After building Archivist for a few years, processing over 10,000 sessions, and working with thousands of TTRPG enthusiasts from casual players to pro GMs, one thing has become very clear to me: there is still no shared agreement on whether AI belongs anywhere near tabletop role-playing games.

That lack of consensus is not surprising. TTRPGs sit at the intersection of creativity, community, and personal expression. People care deeply about them. When something new shows up that feels like it might cheapen, automate, or hollow out that experience, skepticism is healthy, and debates get heated fast.

Our stance: AI should not take the creative reins in a TTRPG. It should protect the creative core by preserving what actually happened.

This guide draws a hard line between admin help and authorship, then shows what that looks like in practice.

A Few Positions, Plainly Stated

• Large language models are currently unreliable Game Masters.
• They are very good at repetitive, unglamorous work that humans tend to resent.
• There are legitimate ethical questions around how modern AI systems were trained and deployed.
• A lot of AI content flooding social platforms feels disposable and exhausting.
• The speed at which AI has entered public life has given people very little time to adapt.

None of those points are controversial on their own. The trouble starts when “AI” becomes a single symbol for a dozen different fears.

When that happens, the debate stops being about specific use cases. It turns into identity and ethics, and everyone talks past each other.

Why the Debate Around AI in TTRPGs Gets Heated

When AI comes up in tabletop spaces, the objections usually fall into a few broad buckets.

1. Some people only know AI through generic chatbot output or endless AI images and videos. It is easy to imagine that kind of content invading a campaign and making everything worse.
2. Others see AI as fundamentally unethical. The reasons vary from artist compensation to environmental impact, labor displacement, cognitive atrophy, and more. The conclusion is the same: opting in feels like endorsing harm.
3. There is a common assumption that "using AI in a TTRPG" means replacing the Game Master, outsourcing creativity, or letting a model make decisions that belong at the table.

Some concerns are well-founded. Others come from conflating very different use cases. Treating every use of AI as morally equivalent makes it harder to set real boundaries.

AI is already embedded across modern life, in medicine, accessibility, fraud detection, search, and logistics. That does not excuse misuse. It does mean blanket moral claims rarely survive scrutiny.

AI can cause harm. Like any powerful technology, it amplifies both good and bad behavior. The difference this time is speed.

The only productive path forward is honest discussion, clear boundaries, and thoughtful application, especially in spaces built on trust and collaboration.

What Archival AI Looks Like in Practice

Archivist is designed as AI campaign management that takes the raw chaos of a session and turns it into structured, searchable, reliable records. Transcripts. Summaries. Character histories. Faction timelines. Canonized rulings. Not new ideas. Not rewritten scenes. Just a faithful record of what already happened. That is an important distinction.

I have personally gone back through Archivist session data and rediscovered moments that would have been completely lost otherwise:

• Throwaway lines that became future plot hooks
• Offhand NPC interactions that later gained weight
• Rulings we would have misremembered months later

That content was preserved because a model remembered what we could not. Continuity issues vanished. Arguments about "what happened last time" stopped eating the first half hour of sessions. No one had to break character to take notes. No one was stuck playing historian instead of participant.

Most importantly, the creative energy at the table increased. With the safety net in place, improvisation felt freer. My attention stayed on players instead of documentation. My ADHD brain could chase ideas without worrying they would evaporate forever.

Under the hood, this means speech-to-text pipelines, careful transcript cleanup, conservative prompting, embeddings tuned for recall rather than invention, and just enough reasoning to be useful without becoming authoritative. It is intentionally unexciting. It is infrastructure, and that is the whole point.

The Bottom Line

If you want AI in your TTRPG, keep it in the admin lane. Use it to capture what happened, keep continuity tight, and reduce bookkeeping. Do not use it to invent canon, make decisions, or steer the story.

The table owns the narrative. AI can earn a seat only by staying out of authorship.