GenAI-using developers say they mostly employ it for 'research and brainstorming'
While discussions about generative AI often focus on its potential to write code, new data suggests that industry professionals are primarily using the technology for much different purposes.
The question of what purpose generative AI serves in game development has been a central topic since the commercial advent of large language models (LLMs) in 2022. While the role of AI in coding has recently dominated many discussions, new data from the GDC Festival of Gaming State of the Game Industry report indicates that it is not the primary use case among professionals.
According to the report, 40 percent of industry professionals use generative AI in the workplace. Among those who do, 73 percent of respondents identified "research/brainstorming" as their number one use case. Additionally, 62 percent reported using it for "data analysis and summarization," while only 56 percent stated they use it for "code generation/debugging." Respondents were permitted to select multiple options describing their use of the tool.
Forces shaping AI adoption
The adoption of generative AI tooling has faced significant resistance from segments of the professional game development community. Previous data has indicated that half of developers are concerned that the technology may lower the quality of games. Throughout recent surveys, respondents have also expressed concerns regarding the environmental impact of AI tools and the use of copyrighted material in foundational training data.
Conversely, some developers report that generative AI has enhanced their programming capabilities. This aligns with comments from experienced programmers, such as DayZ and Rocketwerkz creator Dean Hall, who explained that the tool has accelerated his ability to develop games using complex programming languages and frameworks. This approach is not described as "vibe coding"—the practice of progressively prompting AI to create software features—but rather as leveraging the ability of LLMs to "tokenize language."