itch.io AI Game Policies in 2026: Disclosure Rules, Asset Bans, and What Gets Rejected

Why itch.io's AI Rules Matter More Than Steam's Right Now

itch.io has become the default storefront for AI-assisted indie experiments — partly because it is fast to publish, partly because its review process is human and its fee model is generous. That openness is now being tested. In 2025, itch.io began enforcing a disclosure regime that asks creators to flag AI-generated content at the asset level, not just at the project level. The result is a policy that is easy to comply with if you plan ahead and surprisingly easy to fail if you treat it as a checkbox.

This matters because itch.io is where many developers first ship. A rejected page or a delisted project is not just an inconvenience; it is a signal to publishers and platform reviewers. The developers who are navigating the rules successfully are not avoiding AI — they are documenting it.

What the Policy Actually Requires

Disclosure is asset-level, not project-level

itch.io's guidance asks creators to identify AI-generated content in the project description and, where applicable, in the asset metadata. The distinction matters: a game with an AI-generated soundtrack and hand-painted sprites is not the same disclosure case as a game with AI-generated character art. In practice, reviewers look for a clear statement of which assets were generated, which tool was used, and whether the output was edited afterward.

A common mistake is a blanket sentence like "This game uses AI." That is too vague to be useful to a reviewer and can trigger a manual review request. A better pattern is a short table or bullet list: asset type, tool, model version, and post-processing steps.

The asset categories that draw scrutiny

The categories that attract the most review attention are:

• Character and key art: AI-generated sprites or portraits that appear to imitate a specific living artist or a protected franchise.
• Voice and music: cloned voices and generated tracks, especially where the source model or training data is unclear.
• Code: AI-generated scripts are generally acceptable but should be disclosed if they form a substantial part of the project.
• Store page assets: AI-generated capsule images and screenshots are held to the same disclosure standard as in-game assets.

itch.io does not ban AI content outright. What it rejects is undisclosed or misrepresented content. The enforcement lever is the page review, and the most common outcome is a request for clarification rather than an immediate ban.

The Compliance Workflow That Holds Up

A workable process has three stages: inventory, provenance, and presentation.

Inventory means listing every asset and marking its origin before you start building the store page. This is easiest if you keep a simple spreadsheet from day one. If you are generating 3D props, tools like Tripo3D and CSM can produce engine-ready models quickly, but each output should be logged with its prompt and model version. For 2D art, Krita with AI plugins is a common choice because the plugin history provides a natural audit trail.

Provenance means being able to answer, for any asset, where it came from. This is where most developers fall short. A generated track from Suno or Udio should be saved with the generation date and the model tier. A cloned voice from Resemble AI should be documented with the consent status of the source speaker.

Presentation means writing the disclosure in the project description in plain language. Reviewers are not looking for legal language; they are looking for honesty and specificity.

Real Cases and Data Points

itch.io's policy shift did not happen in a vacuum. In September 2024, Steam began requiring developers to disclose AI-generated content during the submission process, with a pre-generated and live-generated distinction. That framework influenced how smaller platforms approached the problem. itch.io's own guidance, updated through 2025, emphasizes creator responsibility and platform transparency rather than a blanket prohibition.

The practical effect on the ground has been uneven. Developers report that pages with clear disclosure statements pass review without friction, while pages with vague statements are more likely to receive a clarification request. In one widely discussed case in early 2025, a visual novel with AI-generated background art was temporarily unpublished after reviewers found the store page did not mention the art's origin; the developer republished within days after adding a disclosure section.

The tooling side has also matured. By 2026, most major AI art tools include export metadata that can be attached to asset files, and engine-side asset pipelines in Unity and Unreal can carry provenance tags. This makes compliance cheaper than it was two years ago, but only for developers who build the habit early.

Practical Advice for Developers

• Disclose at the asset level. Name the tool, model version, and post-processing steps for each AI-generated asset category.
• Keep a provenance log. A spreadsheet with asset name, tool, date, and prompt is enough. Update it as you generate.
• Avoid style imitation. Do not prompt for the style of a living artist or a protected franchise; this is the fastest route to a rejection.
• Document voice and music consent. If you use cloned voices, keep written consent from the source speaker.
• Treat store page assets as in-game assets. AI-generated capsule art needs the same disclosure as AI-generated sprites.
• Write the disclosure before you upload. Draft it as part of the build, not after a review request.
• Respond to clarification requests quickly. Most issues are resolved by adding specificity, not by removing content.

Outlook

itch.io's approach is likely to stay disclosure-first rather than prohibition-first, because the platform's value proposition depends on openness. The pressure point will be enforcement capacity: as AI-generated content volume grows, manual review will need better tooling, and provenance metadata is the most plausible path. Developers who build documentation habits now will spend less time in review queues later. The ones who treat disclosure as an afterthought will keep discovering that the policy is stricter in practice than it looks on paper.