AI in gaming: innovation with caution
By Sabine Oskamp
AI is revolutionizing many industries, but gaming stands out for a number of reasons. Gaming is where entertainment meets innovation. Where new AI approaches can be tried, tested and refined before being deployed to other domains.
However, game studios are approaching some applications of AI with caution. The challenge lies in striking the right balance — leveraging the potential of AI without using for the sake of it, all while preserving the magic of games.
To get the latest insights, we spoke with four leading experts at different intersections of gaming and AI in the Netherlands:
Laurens Rutten, Founder and CEO of CoolGames and board member of the Dutch Games Association
Johannes Pfau, Assistant Professor at Utrecht University, specialising in the intersection of video games, AI, analytics and data science
Mike Preuss, Associate Professor at Leiden University specialising in game AI research, natural computing and social media computing
Mo Seifi, Co-Founder and CTO at Rebellis AI, developing text-to-motion generative AI for games
What You'll Learn:
How Dutch game studios are using AI to speed up development and enhance creativity
Why the gaming industry makes an ideal testing ground for cutting-edge AI
How researchers are using AI to understand and improve player experiences
The surprising ways gaming AI innovations can transfer to other industries
Where experts see the biggest opportunities ahead
Why gaming is the perfect testing ground for AI
A common theme in our interviews is that the gaming industry is an ideal testing ground for AI innovation. It allows new methods to be tested before they are used in other areas.
This pattern — of technologies developing first in gaming before transferring to other domains — is well established. From graphics processing to physics simulations, gaming has often pioneered technological approaches that later became standard in fields ranging from medicine to architecture.
"Games have always been a testbed for modern AI approaches because they are small and well-defined. You can handle them," notes Dr. Mike Preuss, Associate Professor at Leiden University, specializing in game AI research. This makes games ideal for testing new algorithms and approaches before applying them to more complex real-world challenges.
"In gaming, we can use creativity to test anything, any idea we have," explains Mo Seifi, Co-Founder and CTO of Rebellis AI, a company developing text-to-motion generative AI for games. This sentiment — that gaming provides a sandbox for experimentation — was echoed by all of our experts.
The forgiving nature of games makes them particularly valuable for AI development. As Seifi points out: "You have that freedom, because a bad animation in a game is funny, but in robotics it is a failure and potentially dangerous, right?"
This lower-stakes environment allows developers to push boundaries without serious consequences.
The evolution of AI in gaming
The relationship between gaming and AI runs deep, predating the current AI wave.
"I started working at the intersection of gaming and AI around 2007. But it was small back then, we worked on Pac Man, for example. It was very much the beginning and not that well organized," explains Mike Preuss.
Johannes Pfau shares a similar timeline: "My first contact with AI in gaming was during my undergraduate studies in cognitive science back in 2011. Back then, AI was pretty unexplored in Central European universities, and not really covered by a lot of programs. But I got into some old-school AI courses which involved planning, search and optimization — for what chess computers do, for example."
Any discussion of AI and gaming wouldn't be complete without acknowledging the watershed moment when DeepMind's AlphaGo defeated world champion Lee Sedol in 2016. This breakthrough demonstrated that AI could master a game once thought too intuitive and complex for machines.
This milestone marked a turning point in both public perception and technical possibilities, setting the stage for today's AI revolution in gaming and beyond.
Laurens Rutten, Founder and CEO of CoolGames, puts AI in perspective of other trends that have come and gone: "We've seen a couple of really big hypes in gaming. We had the hype of the metaverse, we had the hype of web3 and crypto. And now we have AI. But I think the difference is those first two were just really hypes and only will be applicable to niche parts of the industry. While AI is something so big, coming so fast, that it's not a hype — it's a reality."
The more recent evolution of AI is visible in how game studios are approaching it today. For many studios, the integration of AI has been gradual.
Rutten outlines four distinct stages of the most recent AI adoption:
"We see four different stages. There was phase one, which started a few years ago until about a year ago, where it was mostly used for making concept art or helping with text. But it was a lot of trial and error back then," Rutten explains. "The big change has come in the last 6-12 months — what I call phase two — where we're really integrating AI into the production process. Artists, developers, game designers — they’re all using AI in different ways. Stage three would be using AI more in the games themselves, and stage four would be using AI to actually help create entire games. But I think stages three and four are still some years away."
The current state of Dutch gaming
The academic perspective on gaming AI has also evolved in the Netherlands, though not always positively. "Five years ago, pre-pandemic, computer game companies were interested in working together with academics on AI," notes Preuss. "I see that decline quite a lot nowadays. It's hard to get into contact and get something running with the games industry because they are very careful at the moment."
This caution stems from market pressures. "The Dutch games industry has been under a lot of pressure for the last two years," explains Rutten. "In the early COVID years, anyone with a little bit of a plan or an idea could build games. And they got paid very well for that. Those days are absolutely over. Now you need really cost-effective, marketing-savvy games. And that's where AI has the potential to help."
Current applications of AI in gaming
1. Development and production
The consensus among the experts we spoke to is clear: AI's most significant impact in gaming is happening behind the scenes in development and production. While experimental gameplay applications show promise, it's the streamlining of creation processes where AI is truly transforming the industry today.
AI is transforming how games are made, with dramatic improvements in efficiency. "What used to take two days for an artist now takes two hours," explains Rutten. "There's still a human needed for the finishing touches, but it's way more efficient than even a year ago."
CoolGames has integrated AI across their entire production pipeline:
Developers use it for coding
Artists use it for animations and graphics
Game designers use it for storylines and wireframes
Marketing teams use it for research and content creation
AI-generated reports on automated player sentiment trends and feedback analysis
"It's an integral part of the production process now," says Rutten. "We have to do this. If you don't do it, left and right, studios and other skilled game developers will be running ahead of you."
The drive for efficiency isn't only about keeping up with competition — it's about survival. "The games market is in pretty bad shape," Rutten notes. "In the last two, three years, a lot of studios closed, or a lot of studios let people go. So there's a price pressure to do more effective, efficient game development."
He clarifies that while players continue to enjoy games, the industry itself faces unprecedented challenges due to intense competition and significantly lower investment budgets for development.
2. Player behaviour analysis and testing
Understanding how players interact with games is becoming increasingly sophisticated through AI.
Pfau, who developed an optimization tool for player improvement, explains: "We had the idea of tracking all the decisions that players are making into an interpretable tool that you can access online. Then we put AI to basically replay what you did, and also branched with classic AI methods to search for things that you could have done differently."
The results were impressive. "We ran a study and found that people who used this for four weeks improved more than those who just played regularly."
This kind of analysis extends to quality assurance as well. Pfau, who developed an automated testing system called Icarus, explains the problem it solved: "Play testing — basically playing the same game to test it, doing the same levels, seeing the same content, and reproducing the same steps over and over again — is not fun for anyone. You lose the fun of playing video games after a while."
His solution was a semi-automated system combining AI agents with human oversight. Icarus wasn't just an academic experiment — it was successfully implemented in industrial applications for about six years, demonstrating its practical value for game development teams. "We could test the game in depth using a speed run method," Pfau notes. "Even more interesting and harder for humans to evaluate is exhausting all the possibilities that could go wrong."
The system worked in multiple modes: "We were able to make the agent avoid making progress before all options were tried."
3. Animation and motion
Some companies are pushing the boundaries of what is possible with AI in gaming. Rebellis AI is focusing on text-to-motion generation. Mo Seifi, Co-founder CTO, explains their approach:
"It's a text-to-motion generative AI. So we make humanoid motions on human characters, and we're very focused on gaming right now, especially indie game studios."
They offer two distinct ways to use their technology:
As a co-pilot for animators: "The first one is as a co-pilot. Any game studio, big or small, has animators that create animations based on the amount of effort they can put into this area. But with our system, instead of having a big team of animators, you can accelerate your animators' work, which is especially valuable for indie game studios."
For runtime animation: "The second is using our system directly in the runtime of the game. It's a use case where the AI works during gameplay — the game developer creates the frame, and players do the prompting or use presets. The creativity and making of the game is based on the players' creativity."
The impact on development speed is significant: "A 10- or 15-second animation that usually takes a week for animators, takes two hours with our approach," notes Seifi.
Their approach also enhances creative flexibility: "As an artist, as a creative, as an animator, sometimes you don't know what you need. You have a general idea, but you don't have that detailed idea yet. So you can use our system to test and prototype."
Successful games and AI applications
While many AI applications remain behind the scenes, our experts shared examples of their work and notable games:
CoolGames has developed several successful and well-known titles including:
Farm Merge Valley on Discord: "It's the number one game on Discord at the moment," notes Rutten. "We've had millions of players since launch, and still have hundreds of thousands of players every day."
Battleship for Netflix Games: A modern version of the classic board-game using HTML5 technology under the hood
However, as Rutten clarifies, CoolGames' AI implementation is primarily in development processes, not in the final shipped games. The company remains in "phase two" of AI adoption. He believes that using AI within games themselves is "still some time away” at CoolGames.
Rebellis AI, on the other hand, has seen their AI animation technology used in some shipped games:
Weeny Wise on Steam: One of the first games to use their text-to-motion AI
Roblox games (in development): Seven games are using their real-time animation technology: "We have seven games that studios are working on integration. Most of them are in Roblox. While they are getting close, none of those have been released yet," notes Seifi.
In the broader industry, Pfau points to different success stories:
Resident Evil 4 (2005) pioneered dynamic difficulty adjustment, though with mixed reception
Modern multiplayer games use AI for matchmaking: "The closest thing to difficulty adaptation that we have nowadays, and that works, is skill-based matchmaking," notes Pfau.
His own Guild Wars 2 analytics platform has thousands of users who improve their gameplay through AI-driven insights
How indie game studios are leading AI innovation
While larger studios often have more resources, our experts suggest that indie game developers are leading the charge in AI innovation.
"Indie studios want to try everything new. They're open to new challenges, new opportunities, new ideas," explains Seifi from Rebellis AI. "Big studios have some success stories and based on them — for example, they have Red Dead 1, and they want Red Dead 2, Red Dead 3..."
The willingness to experiment makes indie studios perfect early adopters of AI. "They're open to any new possible way of using everything. That's what's attractive to us," says Seifi. The efficiency gains from AI tools can be particularly impactful for smaller teams.
The indie space also allows for more creative freedom and experimental applications of AI. "I am a fan of experimental games, wacky games, or games that are trying to do something very different, that might open up new genres," says Pfau. "That is something that, as I already said, not everybody can afford and there's a huge risk at that, but I see that especially the indie games branch can make a lot of use with that and combine that with a lot of crazy and cool ideas."
This experimentation extends to how AI is implemented. "They can use things that you can't do before," notes Preuss. "Like having NPCs actually interact with each other in meaningful ways, or generating content that responds to player behavior in real-time. These are risky features for triple A games, but indie developers can try them out."
The impact of recent AI advances
1. Large Language Models (LLMs)
The gaming industry is carefully watching the potential of LLMs, particularly for non-player characters (NPCs) and narrative elements. However, our experts urged caution.
"I think that's somewhat the low-hanging fruit that everybody thought of when seeing how good large language models are right now," notes Pfau.
He describes a recent student project exploring a novel approach: "Instead of using the LLM for both reasoning and communication, they separated these functions. The LLM handles the communication while a separate system manages game logic. This way, we at least know how the reasoning works."
2. Generative AI
The impact of generative AI is already being felt, particularly in asset creation. "There's a big opening for asset creation," notes Preuss. "2D sprites, artwork with diffusion models, even 3D models, or preliminary models that are manually refined."
However, Rutten emphasizes the need for controlled environments: "If we create art, for example, we do that in a closed environment. Copyright is an issue, and that's still not solved." More on that below.
Challenges and considerations
1. Industry caution
Despite the potential, many game companies remain cautious about implementing AI, particularly in player-facing features. "The industry is, for a lot of good reasons, hesitant when it comes to putting AI at the point where there is an interface to the player," explains Pfau. "It could be very experimental, very wonky, and it could create the wrong experiences that you can't design or influence."
This caution isn't necessarily bad, and not everyone is playing it safe, according to Seifi: "Early adopters are the younger companies. Because indie studios need to be nimble and efficient, they're open to it. Big studios are more stable, so they can wait to see everything tested and made better first."
2. Player perception
There is a risk of AI being perceived negatively by players. "AI is not perceived as exciting by a lot of people in gaming, but as being cheap and unimaginative," notes Pfau. "Because everybody can just throw AI at a game now. For example, you can just prompt two sentences into the backstory of a character. Then some of the magic gets lost."
He notes the risk of diminishing the value of human creativity: "It might not be as valuable as a fully drawn out, fully-fledged dialogue tree with a lot of thoughts from people who are very good in narrative."
3. Market pressures
The gaming industry is facing significant challenges. "You need cost-effective, marketing-savvy game developers now," explains Rutten. This has led to increasing requirements for AI expertise: "I see it in vacancies — people need to have experience with AI, if they are going to be hired. And I think that trend is only beginning."
4. Copyright and ownership
"The whole copyright issue is a big issue," emphasizes Rutten. "Who owns these things that AI creates? This will take a long time to solve." This uncertainty has led to practical adaptations in how companies use AI: "We try to do everything where we have AI helping us in a closed environment, so that's not being part of some other game studio suddenly having it, or it pops up in some video."
5. Jobs and skills
The potential displacement of creative professionals is a concern amongst our experts. "It replaces the jobs of people. I don't know if that's a good thing," Rutten acknowledges. "On the other hand, it will probably also create new jobs. So people will be forced a bit, in some cases, to change their skill sets."
Technical implementation and challenges
1. Frameworks and tools
Our experts highlighted several key tools and approaches for implementing AI in games:
For development:
GitHub Copilot for code assistance
Layer.ai for game-specific art generation
Unity plugins for LLM integration
Local servers running Stable Diffusion through ComfyUI
"We use the same frameworks as people usually use in the rest of AI," notes Preuss. "The overlap is quite strong."
For Rebellis AI's animation system, Seifi describes a three-part architecture:
Pre-processing: "Using LLMs to refine and understand prompts"
Core model: "A vector quantized autoencoder combined with transformer-based models"
Post-production: "Tools for editing pace, transitions, and game-engine integration"
2. Integration challenges
Our experts highlighted several integration challenges beyond just technical implementation. One key issue is compatibility. Many AI tools are optimised for engines like Unity, while support for HTML5 and other ecosystems lags behind. "A lot of tools or plugins based on AI are becoming available quicker to Unity than for HTML5," notes Rutten.
Beyond engine limitations, AI’s demand for high processing power can impact performance, especially for real-time applications like dynamic NPCs or procedural content generation. Latency and computational costs remain obstacles, particularly for mobile and browser-based games.
Another challenge is maintaining creative control. While AI can accelerate asset creation, over-reliance risks producing generic or uninspired content. Studios must strike a balance, ensuring AI enhances rather than replaces human artistry.
Player perception matters too. Some gamers view AI-driven content with skepticism, preferring handcrafted experiences over AI generated ones. Studios must be mindful of how AI impacts gameplay, storytelling and overall immersion for players.
3. Data collection and privacy
When it comes to collecting player data, Pfau emphasizes the importance of consent and control in his own approach: "I only use data that people actually submit on their own end. I ensure that I'm not tracking somewhere in the wild and then having to ask for forgiveness later."
He emphasizes the importance of player control: "Offer options that give full control over player’s data, so you make it inaccessible. You can remove any and all information that is associated with someone’s account."
His platform for Guild Wars 2 analytics demonstrates this approach:
Only uses data that players explicitly submit
Gives users full control over their data accessibility
Allows removal of any associated information
Handles "multiple dozen terabytes of data" while maintaining privacy
Future trends and opportunities
1. Further game development optimization
"Narrowing or closing the gap between ideation and creation" is a key opportunity, according to Pfau. "What a game designer or narrative person has in mind — to communicate that to artists, level designers, people who do scenes and trailers and videos and characters — that needs decades of industry experience. Any effort to close this gap is a good thing."
2. Dynamic content creation
The potential for AI to create truly dynamic game experiences is exciting. "You can have NPCs interacting with each other, dynamic quests, events that are emerging that are not written into a game," explains Pfau. "I find this really interesting. Although I'm not sure how long the appeal will last once you are playing a game that has promised to have infinite content."
3. Video generation
"One technology we haven't used yet, but I think will be super important, is Sora," says Rutten, referring to OpenAI's text-to-video model. "The things that I've seen coming from that are amazing. Even if it's little video scenes between levels — I see a big future for us there."
The impact could be significant: "Until recently, you needed specialised studios for creating video content and animations — very labor intensive and costly."
Of course Seifi and Rebellis AI are already pioneering this.
4. Experimental games
"I am a fan of experimental games, games that are trying to do something very different, and AI might open up new genres," says Pfau. "Especially the indie games branch can make a lot of use of AI and combine that with crazy and cool ideas."
Transferring AI from games to other fields
Gaming has long been a proving ground for AI breakthroughs, but its impact goes far beyond entertainment. The AI techniques pioneered in game development are increasingly being applied to real-world challenges in other industries.
One standout example is how search algorithms originally developed for strategic gameplay are now used in chemistry research. Mike Preuss, Assistant Professor at Leiden University, shares how his work with AlphaGo-inspired AI unexpectedly led to a major scientific breakthrough:
"We applied the AlphaGo algorithm to a chemistry problem, and it worked so well that it was accepted in Nature. Chemists are now actively monitoring AI conferences to see what they can adopt next."
Beyond scientific research, AI innovations from gaming are influencing a wide range of fields:
Search algorithms: first developed for strategy games, now used in fields like logistics, pharmaceuticals, and financial modeling
Game-based learning techniques: applied to professional training, improving education, corporate learning, and even medical rehabilitation
Visualization and interface design principles: Game UI innovations are shaping everything from virtual workspaces to architectural simulations
User engagement, gamification, and motivation strategies: AI-driven player engagement methods are now widely used in marketing, social media, and behavioral analytics
Looking ahead, some of our experts believe the next frontier is robotics. Seifi sees gaming AI’s potential to bridge the gap between virtual and real-world movement: "If we can perfect real-time AI-driven movement in games, why not apply it to robotics next?"
This idea of gaming as more than just entertainment is echoed by Pfau: "There's a lot of inherent skills in video games — training muscle memory, hand-eye coordination, reflexes and decision-making under stress. These could have applications far beyond gaming."
Preuss adds: "Gaming is also great for English learning. Gamers are often much more fluent in English. It's generally good for education as it motivates people. There's lots of untapped potential, and I have seen that other countries, especially Asian countries, are much further with that."
Advice for AI professionals
For AI professionals interested in gaming, our experts offer several recommendations:
Understand the domain "You need to understand what makes a game tick," advises Rutten. "If you're a software engineer, by building a game, you will learn different kinds of software skills than if you are working on some banking program or office application."
Consider multiple paths "The first question is, do you want to develop games for fun? Yes, please do it," says Rutten. "Do you want to live from it? Then you should ask yourself a couple of questions first."
Focus on practical solutions "If you start from the perspectives of what can cut costs, what can produce easy-to-communicate strategies or insights for game developers, then you are on a good track," advises Pfau. "If you're saying 'Hey, I have a crazy idea, and let's build a game around this,' then you would be better off starting an indie company yourself."
Conclusion
AI in gaming is at an exciting juncture. Industry caution and player perceptions are key issues. But the potential is huge. From accelerating development to enhancing the player experience to enabling completely new types of games, AI is reshaping what's possible.
The Netherlands, with its established gaming industry and growing AI ecosystem, is well-positioned to be a leader — but success will require continued collaboration between academia, industry and the AI community.
For AI professionals and game developers, this presents opportunities — but also demands careful consideration of how AI can enhance rather than diminish the magic of gaming.
As our experts have shown, success lies not in using AI for its own sake, but in finding ways to amplify human creativity, improve player experiences and push the boundaries of what's possible in interactive entertainment.
And as it has done throughout computing history, innovations pioneered in gaming will continue to influence AI applications across industries.
