Project Genie is out of the bottle, but it’ll be a while before Google’s AI can make anything real
You can trust VideoGamer. Our team of gaming experts spend hours testing and reviewing the latest games, to ensure you’re reading the most comprehensive guide possible. Rest assured, all imagery and advice is unique and original. Check out how we test and review games here
As of January 29, 2026, Google DeepMind has unleashed its latest world model, Project Genie. This is available to any Google AI Ultra subscribers in the US, which is anyone willing to pay $249.99 (£234.99) a month. This immediately sent a shockwave through the video games industry. Project Genie didn’t just create videos like Sora AI, but full playable environments that should auto-generate as you explore them.
With yet another AI threat to the games industry, Investors immediately got frightened. Stock prices plummeted for Grand Theft Auto 6 owner Take-Two Interactive and celebrated Witcher 3 developer CD Projekt Red. However, it’s a bit early for these investors to be pulling out, because Genie is hardly the death of the industry. At least, not yet.
- Google DeepMind has opened up access to Project Genie, its latest generative AI model for world creation, and this immediately sparked fears among investors.
- Project Genie is only a research model right now, and is restricted to short bursts of content, making the threat to the video games industry unclear.
- The AI tool isn’t actually able to inject any gameplay into these worlds, making them playable only in a very loose sense.
- Project Genie might also run into the issue of copyright, as it seems very good at generating knock-offs.
- Most of all, Project Genie is of little threat to the video games industry because consumer opinion is powerfully weighted against AI.
Short and sweet
Project Genie is impressive at first glance. It’s easy to see why investors might be spooked. It’s a perfect example of AI technology potentially replacing real artists in the video game industry. Investors look at Project Genie and see a world where anyone can come up with an idea and suddenly find themselves exploring a new world like a Star Trek holodeck.
However, investors have pushed the button a bit early. Right now, Project Genie can only generate experiences that are at most a minute long, and capped at 24fps and 720p as well. This is perfect for generating short eye-catching videos made for marketing, but it’s not helpful when making a game.
Where it could be useful, however, is in helping companies develop worlds. Recently, Sven Vincke, head of Larian Studios, spoke about using AI to develop concept art. Project Genie’s ‘world remixing’ is exactly what Vincke would want out of this, being able to plug in a concept and use AI to iterate for a relatively small investment of time.
However, with its current minute-long restrictions, Project Genie could only be a threat to games like fast-paced mini-game collections like WarioWare. Even this is scuppered, however, since Project Genie can’t actually generate gameplay.
Games often need play, actually
Project Genie is a world builder, but its ability to understand physics has pushed people to see game-making potential that isn’t there. You can walk around, you can glide, you can even jump and collect coins in a fake Subway Surfers environment.
However, this all amounts to less interactivity than ‘walking simulators’ like Dear Esther. Project Genie engages in ‘world sketching’, which is just a slightly interactive video. So, you’re able to make the game world and wander around as Project Genie generates terrain in your vision cone, and often forgets what it generated behind you.
If you were able to permanently extrude these environments in some way, maybe then you’d be able to start putting actual gameplay in. You could make a point to the coin collections, put some enemies out there, create a failstate, or something else that entertains for more than a minute.
Except, at that point, you’re just making the game anyway, but with the equivalent of a pre-made Unity asset background. This isn’t anything new; train horror game Choo Choo Charles was made almost entirely with stock Unity assets, and, while it found little critical success, it managed to get at least some time in the limelight.
This step into making actual entertainment still requires game developers, which these investors appear to have forgotten when selling off their stock.
Rote learning and regurgitation
Generative AI like Project Genie can make a lot of things, but it can only work with whatever knowledge it has to hand. This makes it very good at generating environments based on real life, or on games where there’s a lot of footage floating around, like The Legend of Zelda Breath of the Wild.
However, generative AI cannot imagine; it can only remix. AI is the worst kind of people pleaser, and it likes to regurgitate things that are popular because that’s how it gets some of the best feedback. So, if developers want something standardised, they’re onto a winner.
Right now, copyright is an unresolved issue with generative AI, as companies that would normally be quite protective of their IP, like Disney or Microsoft, are actively investing in AI and thus being more lenient.
When it comes to full-on game generation, it’s not yet clear how distinct assets accidentally reclaimed from the AI slop will be defended in court. Certainly, Project Genie is waving a red flag at the bull that is Nintendo, which is well known for striking fan projects like Rhythm Heaven Remix.
There’s not even much that generative AI can do about this. It has learned to use copyrighted content, and when encouraged to make more, will start learning from its own models. At best, Project Genie can do something that’s already been done, and at worst will leave companies with a ticking legal time bomb.
Nobody wants this
Perhaps the biggest reason that game spaces aren’t quite ready to be taken over by AI is simply that players don’t want AI in their games. After the recent backlash, Larian Studios has confirmed it isn’t going to use any “GenAI art in Divinity”, or even during “concept art development”.
Take Two Interactive’s CEO, Strauss Zelnick, has also said that although “generative AI tools are driving costs and time efficiencies”, that “with regards to GETA 6, generative AI has zero part in what Rockstar Games is building.”
He reiterates, as Vincke has said previously, that AI is being implemented for “mundane tasks”. This tempered reaction to generative AI is certainly reeling from overall negative impressions from within and without the industry.
In a survey from Quantic Foundry, nearly 63% of consumers had a very negative opinion of AI in video games, and in GDC’s State of the Industry Report, 52% of game development professionals think AI is hurting the games industry.
Shareholders who have no idea what any given market entails can get cold feet at the smallest of moments. As of February 5, 2026, Take Two Interactive’s stock is still 17% down on where it was before the Project Genie announcement, yet Project Genie is still a long way off from ever producing anything of worth, beyond fodder for sizzle reels.
FAQs
What is Project Genie?
Project Genie is the latest world-building model from Google DeepMind, now open for high-paying AI enthusiasts to play around with.
Is Project Genie free?
No, Project Genie is locked behind a Google AI Ultra subscription, and you can only access this if you are in the US. It currently costs $249.99 (£234.99) a month.
What’s the difference between AI and Gen AI?
AI is a very vague term, but it refers to the idea of using machines to analyse data, usually with parameters set by humans. Generative AI, such as Project Genie, is focused on creating wholly new content out of existing content, in particular images, and has a stronger leaning on machine learning.
Subscribe to our newsletters!
