itch.io AI Game Policies in 2026: A Platform-by-Platform Upload Guide

Why itch.io's silence on AI is the real story

Most storefronts have now answered the AI question in public. Steam requires developers to disclose pre-generated and live-generated AI content in a dedicated form, a policy that went live in January 2024 and has been amended several times since. Apple's App Store review guidelines require disclosure of generative AI content and age-appropriate labelling. Google Play added its own AI-generated content policy in 2024. itch.io, by contrast, has never published an AI-specific rule page.

That absence gets read two ways, and both readings are wrong. The first camp assumes itch.io is a free-for-all where anything goes. The second assumes the platform is quietly hostile to AI and will bury AI-heavy projects. Neither matches how itch.io actually operates. The platform has always governed by a small set of broad principles — no hateful content, no doxxing, no malware, no undisclosed asset flipping — and applies them case by case. AI content falls under those existing rules rather than a new one.

For indie developers in 2026, that means the compliance question is not "what does the AI policy say" but "which existing itch.io rules does my AI pipeline touch, and can I prove I followed them." That is a documentation problem more than a legal one.

What itch.io actually enforces

The rules that matter for AI-generated content

itch.io's content rules and quality guidelines are short. Three of them do most of the work for AI projects.

First, the platform prohibits content that infringes third-party rights. This is the clause that catches AI-generated art, audio, and text produced by models trained on scraped data without clear licensing. itch.io does not adjudicate training-data legality itself, but if a rights holder files a credible complaint, the project comes down. That is the same posture Steam takes, and it puts the burden on you to know where your assets came from.

Second, itch.io expects project pages to accurately describe what the buyer gets. If a page sells itself on hand-drawn art and the art was generated, that is a misrepresentation problem regardless of any AI policy. This is where most real enforcement actually happens — not on AI use itself, but on pages that mislead.

Third, the platform's stance on asset flips and low-effort uploads predates generative AI by years. A project assembled from unmodified AI outputs with no design work behind it is more likely to be flagged as spam than as an AI violation. The distinction matters: itch.io is not policing your tools, it is policing whether the result is a real game.

How enforcement actually reaches you

itch.io does not run automated AI-detection scans on uploads. There is no classifier flagging generated images the way some stock platforms do. Enforcement is reactive: reports from users, rights-holder complaints, and manual review of pages that trip spam heuristics.

In practice, that means your risk profile depends heavily on visibility. A quiet jam entry with 200 downloads is unlikely to attract scrutiny. A project that trends on the front page, gets covered by a YouTube channel, or sells well will get looked at. If you are planning a commercial launch on itch.io, assume someone will ask where the art came from, and have an answer ready.

What the 2025–2026 indie landscape shows

The clearest signal from the past year is that disclosure is becoming a marketing asset rather than a liability. Several high-profile itch.io releases in 2025 shipped with a short "AI tools used" section in the project description — usually naming the specific tools, the parts of the pipeline they touched, and what was human-authored. That practice started as risk mitigation and has become a trust signal for buyers who are tired of asset-flip storefronts.

The tooling side has matured too. Teams building AI-assisted art pipelines now commonly run generated assets through cleanup stages rather than shipping raw output. A typical pipeline might generate a base image, then push it through Adobe Photoshop (AI) for cleanup and compositing, or convert it to clean vectors with Vectorizer.ai when the target is UI or iconography. For pixel art projects, Piskel remains a common finishing step because generated sprites almost always need manual frame alignment.

On the audio side, the same pattern holds. Developers generating music with tools like Soundraw or Suno increasingly document which tracks were generated and which were licensed, because music rights complaints are the fastest route to a takedown on any platform.

Two concrete data points frame the stakes. Steam's AI disclosure policy, live since January 2024, now covers well over 1,000 shipped titles that have declared generative AI use — a figure Valve has referenced in developer communications and that has grown steadily through 2025. And in the EU, the transparency obligations under the AI Act began applying to general-purpose AI systems in August 2025, which pushes disclosure expectations upstream to the model providers and, indirectly, to anyone building on them. itch.io has not mirrored either regime, but the direction of travel across the industry is unambiguous.

Practical advice before you upload

• Write an AI disclosure section into your project page. You do not need a legal template. Two or three sentences naming the tools, the assets they produced, and what you changed by hand covers the realistic risk.
• Keep a provenance log. For every generated asset, record the tool, the date, the prompt or input, and the licence terms of the model or service. If a complaint ever lands, this is the only thing that resolves it quickly.
• Do not describe generated work as hand-made. This is the single most common cause of takedowns, and it is entirely avoidable.
• Clean up raw output before shipping. Untouched generations read as low-effort and trip spam heuristics. Post-processing is also where the actual design work happens.
• Check the licence of every model and service you use. Some restrict commercial use, some restrict use in games specifically, and some have changed terms mid-project. Re-check before launch, not after.
• Keep a human-authored core. A project where the design, code, and level layout are yours and AI assisted on assets is a very different risk profile from one where AI produced the whole thing.
• If you are unsure, ask itch.io support before launch. The team answers. A pre-emptive question is far cheaper than a post-launch dispute.

What to expect through 2026

itch.io is unlikely to publish a dedicated AI policy in the near term. The platform's identity is built on being permissive and low-friction, and a formal AI rulebook would cut against that. The more probable path is what we are already seeing: continued case-by-case enforcement under existing rules, with disclosure emerging as a community norm rather than a platform mandate.

The pressure will come from outside. As the EU AI Act's transparency provisions work through their implementation timeline and as rights-holder litigation over training data continues, platforms that stay silent inherit more risk. At some point itch.io will need a position, even if it is a minimal one.

For developers, the practical takeaway does not depend on which way that goes. Document your pipeline, disclose honestly, and make sure the game is actually a game. Those three things keep you compliant under itch.io's current rules and under almost any policy the platform adopts next.