The effects that AI-generated art is having on the development of new games

Art generated with artificial intelligence is no longer a rarity in the video game industry. It has gone from being a laboratory curiosity to a production tool for building prototypes, testing visual styles, creating temporary assets, and cutting the cost of entire stages of pre-production. But that time-saving does not come alone: it also brings doubts about rights, quality-control issues, rejection from part of the audience, and an increasingly bitter debate within studios themselves about what is gained and what is lost when part of the artistic process no longer passes through human hands.

From the instant sketch to the playable prototype

The first major effect of AI-made art is speed. GDC’s State of the Game Industry 2026 survey points out that more than a third of industry professionals already use these tools in their work, and that one of the most common uses is prototyping. That fits with what is being seen in practice: AI is not entering first through the game’s final cover art; it is now present in the earlier stages. It has moved to the point where, in the past, weeks of concept art, interface variations, or environmental tests were needed to decide whether an idea was worth keeping alive. This should accelerate the development of slot games quite a lot, as it can help developers try different artistic styles before continuing with the game itself.

The trend is also moving toward tools that are more integrated into the pipeline itself. In February 2026, Roblox launched a technology capable of generating functional models inside the game itself from natural-language instructions. Last year, Microsoft developed Muse, a model designed to generate visuals and actions for a video game in collaboration with Ninja Theory. All of this points to a scenario in which AI becomes a production layer connected to design, animation, logic, and testing.

For development teams, especially indie ones, the change can be hugely significant. One example is The Roottrees are Dead. The first version of this title moved forward with AI-generated images because the project did not have a large enough budget to hire traditional visual production from the outset. AI was the entry lever: it allowed an idea to exist before there was enough money on the table to present it properly. The immediate effect on development is clear: more prototypes, more internal demos, and more projects that get past the initial barrier of “we can’t afford this right now.”

The problem begins when the sketch wants to stay in the final game

The second major consequence is that what works to get a project started no longer works to finish it. Steam already requires developers to declare whether they have used generative AI during development or within the final product, and it distinguishes between “pre-generated” and “live-generated” content. In addition, Valve reminds studios that this content cannot be illegal or infringe copyright and that, in the case of AI generating material in real time, the safety barriers must be explained in order to avoid problematic results. In other words, distribution now treats AI art as something that can affect approval, commercialization, and user trust.

Added to that is the legal front. The United States Copyright Office reaffirmed in 2025 that human authorship is an essential requirement for copyright protection in that country. It also concluded that prompts, by themselves, do not give the user enough control to automatically make them the author of the result. At the same time, the agency clarified that using AI as a support tool or integrating AI-generated material into a broader work, as long as there is sufficient human contribution in the selection, coordination, arrangement, or modification, does not prevent protection of the whole. The practical reading is fairly straightforward: AI can speed things up, but it does not eliminate the need for artistic direction, editing, and a human touch if what is wanted is a defensible and coherent product.

Less manual work, more artistic direction, and a battle over style

The third effect is creative and labor-related. The same GDC 2026 survey shows that more than half of industry professionals believe AI is having a negative impact on the industry, and the rejection is even stronger among those working in visual and technical art. That suggests the debate is about aesthetics and the kind of employment this shift is encouraging. For now, many artists are afraid of seeing AI enter precisely the tasks that once helped them find their footing, build experience, and develop a style within a studio.

However, the real shift of work seems more complex than simple total replacement. In several cases, AI is shifting the effort toward supervision, correction, and visual unification rather than the manual creation of every image. The Roottrees are Dead serves again as an example because its commercial version did not succeed by keeping the generated art: they had to produce nearly 40 illustrations over the course of about a year to replace it. This case shows the new division of roles the industry is moving toward, where AI creates a quick reference, but consistent visual identity, the legality of the characters, and the feeling that everything belongs to the same world still depend on an author making careful decisions.

Illustration is not the only area of the industry where there is tension. Voice actors and motion-capture artists in video games have also made demands in the past asking for protection against the use of AI. The message is very similar: the more these tools expand in production, the more pressure there is to set limits on attribution and the replacement of creative work. AI enters studios as a cost-saving solution, but it forces a renegotiation of which parts of the process remain human for artistic reasons, for legal security, or simply because of public perception.

All in all, the impact of art made with artificial intelligence on new games already seems irreversible, although not in the triumphalist form in which it was sold at first. It is shortening start-up times, multiplying prototypes, and giving breathing room to teams that previously could not afford certain stages of development. But it is also pushing studios toward a model in which transparency matters more, artistic direction matters even more, and the question is whether it is really worth keeping an image generated in seconds when it is necessary to answer for its quality, authorship, and effect on the identity of the game.