Foundry VTT Bans AI-Generated Content, Dividing Creator Community

Foundry Virtual Tabletop's new AI Content Policy governs usage of generative AI for packages featured on the official Foundry VTT website and the FoundryVTT.store marketplace. The policy, with its most recent substantive change dated March 18, 2026, drew immediate reaction from module authors and the broader tabletop community.

The core rule is straightforward: prepared content must be human-made, meaning text, images, audio, and marketing materials that are intentionally pre-created must originate from human creative work, though AI tools may provide limited assistance with editing or post-processing. The knock-on effect for creators is significant. The policy applies only to packages distributed through the official listing and marketplace, and as a Foundry VTT license holder, authors retain the right to create and distribute non-compliant packages through channels other than the official platforms. In other words, you can still sell an AI-assisted module, but it will not appear on the official database or store.

Improvised content may still be AI-generated: content produced at runtime in direct response to end-user prompts is permitted, and Foundry VTT states it does not police what users choose to generate at their own table. That carve-out preserves tools that function as live AI interfaces during play, while drawing a hard line at pre-packaged AI assets bundled into a module before it ever reaches a player's screen.

Software code may be AI-generated under specific conditions: package authors must personally understand and be prepared to maintain all code. That requirement triggered some of the sharpest community pushback. Commenters on Enworld argued the rule is nearly impossible to police in practice, with one writing that "not allowing LLM code is currently impossible due to how most programmers are trained, things like code completion also use LLM for example." Another described the requirement as simply "difficult to enforce imho."

On the other end of the spectrum, packages that exclusively contain user-facing content developed entirely without AI tools may designate themselves as "Zero AI," advertising their work as 100% human-created. That label has itself become a point of contention, with at least one community member calling it a misplaced priority: "I find it insane that the FVTT developers prioritize a non-AI label over adding labels for translations, etc. and means to filter out certain results already."

For creators already listed on the platform, if a substantive policy change causes a previously approved package to become non-compliant, a 180-day grace period is afforded during which that package may remain listed. The initial publication of this policy counts as such a substantive change, meaning currently listed content has 180 days to become compliant, after which non-compliant packages may be archived or de-listed.

Foundry VTT acknowledged the split directly in its Discord announcement: "For some of you, this policy will be welcome; reinforcing your own values or providing clarity around acceptable usage. For some of you, this policy will be frustrating or damaging to the way that you would prefer to operate."

The policy carries added weight given that Foundry VTT holds an official partnership with D&D as of February last year, a relationship that puts it in direct philosophical opposition to Hasbro CEO Chris Cocks, who has been openly enthusiastic about AI in tabletop. For the D&D tables running their campaigns through Foundry, the practical impact depends entirely on which modules they rely on: those built on pre-generated AI art or text now face a countdown to compliance or removal from the official ecosystem.