House House confirmed on November 10 that large language models had been used in a limited capacity to address technical issues and bugs during the development of "Treading Deeply." Following the announcement, negative reviews for the game increased on Steam, but the number of positive reviews on November 10 and 11 still exceeded the negative reviews for those respective days.
The team stated that the game would run at half speed for about 20 seconds during sunrise and sunset due to an issue with third-party code, which they were unable to resolve even after seeking help from engineers. The developers then significantly modified a solution provided by a large language model and adopted it. Officials claimed that AI-related code accounted for less than 0.1% of the total codebase, and generative AI was not used for any other purposes.
On the day of the disclosure, 366 negative reviews appeared on Steam, with another 130 added the following day. Some players believed that even if the game itself is enjoyable, consumers should decide for themselves whether to support companies using AI, and therefore developers should disclose such information before purchase.
The number of positive reviews during the same period were 426 and 652, respectively. Supporters argued that the usage was minimal and limited to debugging, making the concentrated negative feedback an overreaction. Other players questioned whether House House's previous stance against the harms of generative AI was consistent with their actual practices, while still acknowledging the quality of the work. The discussion also extended beyond the game itself to cover development tools, transparency of information, and whether Steam reviews are an appropriate platform for expressing ethical positions.