Itch.io Deletes Terabytes of AI-Generated Content and Adds a 'No AI' Search Filter

"Just deleted a couple terabytes of AI slop," the store posted on November 2, according to reports by IGN and FRVR. In another post it said that being able to upload 10 terabytes of poorly optimised AI-generated models did not mean creators should, which led some outlets to report the deletion as 10 TB. Itch.io has not said how many pages or projects it removed.

On November 3, the company said it was adding generative AI disclosure filtering to "a few more places on the site where it was critically lacking", including the search results page, Shacknews reported. Users can now choose a "No AI" option to drop disclosed generative AI work from results, or filter by specific uses such as art or sound.

The filter depends on creators disclosing their own AI use, so projects without a disclosure are not verified as AI-free. Games on the store have had to carry generative AI disclosures for about two years, and itch.io now says pages for tabletop games, books, comics and soundtracks must also complete an AI disclosure before they can be created or edited, according to FRVR. The store's existing quality guidelines allow delisting and bans for inaccurate disclosure and for uploading excessive amounts of automatically generated content.

Itch.io is not banning AI outright. It asks creators not to submit low-quality, unpolished or broken AI-coded projects, particularly in bulk, which it treats as spam.

The clean-up follows a sharp rise in uploads. In an earlier statement, the store said the number of game pages created each day had doubled over the past two months, while purchases and downloads had not.

The move adds to signs that generative AI is straining moderation on digital storefronts. Developers have separately complained that review delays on Valve's Steam are pushing indie games past their launch dates, which some suspect is linked to a rise in AI-generated submissions. For platform operators, the itch.io approach of disclosure plus user-controlled filtering, rather than a ban, is one model for handling a surge in low-cost content without excluding legitimate AI-assisted work.