itch.io AI Game Policies in 2026: What Developers Need to Know Before Uploading

Context: why itch.io's AI stance matters now

itch.io has become the default home for experimental, small-team, and solo-developed games. That makes its AI policy unusually high-stakes. When Steam tightened disclosure requirements for generative AI content in 2024 and began rejecting some submissions, a wave of developers looked to itch.io as a lower-friction alternative. The platform's own guidance has evolved in response — not into a ban, but into a disclosure-and-moderation regime that rewards transparency and punishes ambiguity.

The practical problem is that itch.io's rules are short, distributed across help pages and community posts, and enforced by a small moderation team. Developers who assume "itch.io doesn't care about AI" are the ones most likely to get a page taken down mid-launch. This guide walks through what the platform actually asks for, how enforcement tends to work, and what a compliant upload looks like in 2026.

Core analysis: what itch.io actually requires

Disclosure is the load-bearing requirement

itch.io does not prohibit AI-assisted games. Its content guidelines and creator-facing help material ask that generative AI use be disclosed in a way buyers can see — typically in the game description, the project's metadata, or both. The platform's moderation is reactive: reports from users and spot checks trigger review, and a missing disclosure is treated as a trust problem rather than a technology problem. In practice, that means the disclosure needs to be findable without opening a devlog. A single line in the description — "Contains AI-generated character portraits (Midjourney) and AI-assisted code (GitHub Copilot)" — is the minimum viable version.

Asset provenance is the second front

Because itch.io hosts downloadable builds, moderators and buyers can inspect actual files. The most common failure mode in 2025–2026 has not been "you used AI" but "you claimed you didn't." Pages that describe assets as hand-painted when metadata or obvious generation artifacts suggest otherwise have been the ones pulled. This is why provenance matters more than policy wording: keep a simple asset log per project — tool, version, date, and what was generated versus edited. If a takedown happens, that log is your appeal.

The store page is part of the compliance surface

itch.io's product pages are flexible, which cuts both ways. You can embed AI content warnings, tag projects, and use the description field freely. But the same flexibility means there is no enforced field for AI disclosure — you have to design it into the page. Treat the description, screenshots, and any embedded trailer as a single disclosure unit. If AI voices appear in the trailer, say so near the trailer, not only in a footer.

Real cases and data points

• Steam's 2024 disclosure wave. Valve's January 2024 policy change required developers to describe pre-generated and live-generated AI content during submission. The resulting friction pushed a measurable share of small developers to dual-publish on itch.io. itch.io's response was not a ban but clearer guidance that disclosure must be visible to buyers.
• The 2025 itch.io moderation tightening. Through 2025, itch.io updated its content guidelines and moderation messaging, emphasizing that creators are responsible for the legality and provenance of uploaded content. Reports from developer communities indicate takedowns clustered around undisclosed AI assets and copyright complaints rather than AI use itself.
• Tool-level provenance is now standard. Adobe Firefly ships with Content Credentials, and tools like Photoshop (AI) embed provenance metadata by default. That metadata is increasingly what moderators and buyers check. Using tools that log provenance — and preserving those logs — is a cheap insurance policy.
• Audio and voice carry extra risk. Voice cloning tools such as Play.ht and Resemble AI make it trivial to generate dialogue, but cloned voices of real people are a separate legal problem from AI disclosure. itch.io's guidelines on impersonation and rights apply regardless of how the audio was made.

Practical advice: a pre-upload checklist

1. Write the disclosure line before you write the store page. One sentence naming the tools and the content types. Put it in the description's first screen, not the footer.
2. Keep an asset log. For each project: tool name, version, date, and whether output was used raw or edited. A spreadsheet is enough.
3. Prefer provenance-preserving tools. Firefly-based workflows, Content Credentials, and tools with export metadata reduce the burden of proof if a dispute arises.
4. Separate AI disclosure from rights clearance. Disclosure says "AI was used." It does not grant rights to a cloned voice, a trademarked character, or a scraped dataset. Handle both.
5. Audit your screenshots and trailer. If AI-generated art or voice appears in promotional material, disclose it there too — that is where buyers look first.
6. Respond to reports quickly. itch.io's moderation is reactive. A fast, factual reply with your asset log resolves most disputes before they escalate.
7. Don't rely on "everyone does it." Enforcement is uneven, but the downside — a pulled page during a launch — is concentrated on you.

Outlook

itch.io is unlikely to ban AI-generated games outright; its identity depends on hosting experimental work, and a ban would be unenforceable at its scale. The more probable trajectory is a continuation of what we saw in 2025: clearer disclosure expectations, heavier reliance on user reports, and growing buyer sophistication about provenance. Developers who build disclosure into their workflow now will find the policy cost close to zero. Those who treat it as an afterthought will keep discovering it at the worst possible moment — the launch window.