Platform rules on AI content: Steam, itch, Epic, consoles

workingShip it safely7 sourcesChecked: 2026-08-21

Every claim on this page was checked against the platform’s own documentation or announcements on the checked date above. These policies have been rewritten more than once in the past two years, so treat the checked date as the boundary of the guarantee and confirm on the platform’s live page before you submit.

What Steam requires

Steam is the strictest of the PC storefronts, and its system has two halves.

Pre-generated AI content is anything made with AI tools during development that ships in your game: art, audio, writing, code that players consume. You describe it in a free-text box in the Steamworks content survey, and Valve evaluates it the same way it evaluates all non-AI content. Your description appears publicly in an “AI Generated Content Disclosure” section on your store page, and you can edit it at any time.

Live-generated AI content is anything your game creates with AI while it runs, like an LLM-driven NPC. For this you must additionally describe the guardrails that stop it producing illegal content, because Valve cannot pre-review output that does not exist yet. Players get a reporting tool in the Steam Overlay for flagging illegal or inappropriate AI output, and inadequate guardrails risk removal from the store.

Two boundaries worth knowing. In January 2026 Valve narrowed the disclosure scope: AI tools used purely for development efficiency, like code assistants and transcription, no longer trigger disclosure as long as their output is not consumed by players, as reported across trade press when the form changed. And one restriction has no approval path at all: Valve states it does not want to ship live-generated AI adult sexual content at this time.

Practical consequence: if your game has a runtime AI feature, write the guardrail description before you fill in the survey. Response filtering, topic fences, and length caps are exactly what the box is asking about; the talking NPCs article covers building them.

What itch.io requires

itch.io runs a disclosure and tagging system, launched in November 2024 and unchanged in substance since. The important asymmetry: tagging is mandatory and strictly enforced for assets sold on itch, because of the legal ambiguity around AI asset rights, but for games it is encouraged rather than required.

Disclosing applies an “AI Generated” tag with sub-tags for Graphics, Sound, Text and Dialog, and Code. Tagged projects surface on a dedicated “made with AI assistance” browse page, and projects that declare no AI use get a “No AI” tag that works as the opt-out filter for players who avoid AI content.

Classic game AI is explicitly exempt: pathfinding, behavior systems, procedural generation, dynamic difficulty, and similar self-contained algorithms need no tag. Two hard lines to respect: AI-generated imagery designed to resemble photographic content of real people is banned in adult content regardless of disclosure, and the quality guidelines treat floods of minimally-curated AI output as spam.

Epic: nothing on the store, rules on Fab

The Epic Games Store has no AI disclosure requirement. Its content guidelines contain no AI language, and Epic’s leadership has said publicly that it does not police how developers make their games and opposes mandatory AI labels.

Fab, Epic’s asset marketplace, is the opposite in writing: since May 2025 every asset made with generative AI requires a mandatory “Created with AI” declaration at publish time, and misdeclaring violates Fab’s terms. Sellers can also apply a “NoAI” tag to opt their assets out of AI-training use. Worth knowing as a buyer: trade press and developer forum reports through 2025 found large volumes of undeclared AI assets on Fab, so treat the tag as a signal, not a guarantee, and apply the care detector to anything you buy.

Consoles: one public rule, two closed doors

Xbox has the only concrete public console rule. Microsoft Store Policies section 11.16, “Live Generative AI Content”, announced September 2025 and effective October 14, 2025, explicitly covering the Xbox Store, requires products with runtime generative AI to disclose it in store metadata and in Partner Center at submission, keep the AI’s output compliant with all other content policies, and give players an in-product way to report problematic AI output. If you are planning a console port with a live AI feature, that in-game reporting mechanism is a real engineering requirement, not paperwork.

PlayStation publishes no AI-specific submission policy; whatever requirements exist live inside the NDA-gated partner program. Nintendo likewise has no public policy; its only public statement is that IP infringement gets enforced the same whether or not generative AI was involved. For both, plan to ask your platform contact directly once you are in the partner program, and assume disclosure questions are coming.

Mobile, briefly

Google Play has an explicit AI-Generated Content policy: you are responsible for what your app’s AI outputs, and specific categories are banned outright regardless of disclosure, including non-consensual sexual deepfakes and impersonation for scams or election deception. Apple regulates the data side: App Review Guideline 5.1.2 was reported in late 2025 to require clear disclosure when personal data, including prompts, is sent to a third-party AI service, and AI chat features are reported to fall under the user-generated-content rules that require moderation and reporting tools. If your game calls cloud AI on mobile, the player data article covers what that means in practice.

The cross-platform checklist

Write these once and you have answered every store’s question:

A one-paragraph inventory of AI-generated content that ships in the game, by type. This is your Steam survey text, your itch tags, and your answer when a console platform asks.
A one-paragraph guardrail description for any runtime AI feature: what filters, fences, and caps stop it producing illegal or off-policy output. Steam and Xbox both ask for exactly this.
An in-game reporting path for live AI output if consoles are in your future. Xbox requires it today; the Steam Overlay provides it on PC.
A note of which cloud services receive player input, for privacy disclosures. See player data and cloud AI.

Going deeper

The pattern across every platform is the same shape: disclosure of what ships, accountability for what generates at runtime, and zero tolerance for illegal output wherever it came from. Policies converge on that shape from different directions and at different speeds, which is why the checklist above is written against the shape rather than any one store’s current form. When a policy shifts, it is nearly always one of three moves: scope narrowing (Steam in January 2026), a new runtime rule (Xbox in October 2025), or enforcement tightening on an existing rule. Skim the platform’s developer news feed at each submission, and read Reading tool terms before you ship for the same discipline applied to your tools.

Bring your platform’s specific disclosure question to the AI Game Dev Org Discord before you submit; someone there has probably already fought that exact form. Plenty of the games on the showcase made it through that process first.