TL;DR
Crystal Dynamics and Flying Wild Hog used generative AI to create placeholder assets during pre-production of Tomb Raider: Legacy of Atlantis, accelerating visualization and iteration. The move highlights a growing industry shift toward using AI for non-final, internal development work rather than replacing human artists outright.
What Happened
Crystal Dynamics and Flying Wild Hog have confirmed that generative AI was used to produce placeholder assets during the early development of Tomb Raider: Legacy of Atlantis, a new entry in the iconic action-adventure franchise. The studios told Eurogamer that the tooling helped them "get the right answers faster" by allowing designers and artists to quickly visualize concepts before committing to final, hand-crafted assets — a pragmatic use of AI that sidesteps the more controversial applications seen elsewhere in the industry.
Key Facts
- Crystal Dynamics and Flying Wild Hog co-developed Tomb Raider: Legacy of Atlantis, which was announced in March 2026 and is slated for a 2027 release.
- The studios used generative AI to create placeholder assets — including environmental mock-ups and character silhouettes — during the pre-production phase.
- The AI-generated assets were never intended for final release and were replaced by hand-authored art before the game entered full production.
- The tooling was described as helping the team "get the right answers faster" by enabling rapid iteration on visual concepts without waiting for full asset creation.
- Flying Wild Hog, best known for the Shadow Warrior reboot series, joined the project as a co-developer in 2024.
- The announcement comes amid widespread industry debate over AI in game development, with major studios like Square Enix (Crystal Dynamics' parent company) and Ubisoft having publicly explored generative tools.
- The game marks Lara Croft's first mainline console release since 2018's Shadow of the Tomb Raider, a six-year gap that has built significant anticipation.
Breaking It Down
The key distinction here is that Crystal Dynamics and Flying Wild Hog used generative AI for internal pre-production — not for final, shipped content. This is a materially different use case from, say, using AI to generate dialogue or final textures that players will see in the finished product. Placeholder assets, often called "greybox" or "blockout" art, are typically rough geometric shapes and low-resolution textures used to test level layouts, camera angles, and gameplay flow before artists spend weeks on high-fidelity models. Speeding up this phase with AI-generated imagery can compress the iteration cycle from days to hours.
"The single most revealing figure is that the studios explicitly stated the AI assets were 'never intended for final release' — a defensive framing that underscores how sensitive the issue has become."
Crystal Dynamics is acutely aware of the backlash that followed other studios' AI experiments. In 2024, Square Enix itself faced criticism for using AI-generated promotional art for Foamstars. More recently, Activision and Ubisoft have both faced player revolts over AI-generated in-game content, from Call of Duty loading screens to Assassin's Creed NPC dialogue. By positioning this as a pure pre-production tool — and emphasizing that human artists still create the final assets — Crystal Dynamics is trying to have it both ways: benefiting from AI's speed without absorbing the reputational damage.
The choice of Flying Wild Hog as co-developer is also notable. The Polish studio, which built its reputation on fast-paced first-person shooters like Shadow Warrior 3 and Evil West, brings a different design sensibility to the Tomb Raider franchise. If the AI placeholder workflow allowed the two teams to align on visual direction more quickly — especially given the logistical challenges of cross-studio development — that could be a genuine productivity gain. The question is whether such gains will translate into a better game, or simply a faster one.
What Comes Next
The immediate trajectory depends on how the industry — and players — react to this specific implementation. Crystal Dynamics has been transparent about the tooling, which may inoculate it against some criticism, but the broader AI debate shows no signs of cooling.
- Full production reveal (Late 2026): Expect Crystal Dynamics to show the first real gameplay footage of Tomb Raider: Legacy of Atlantis at a major event like Gamescom or The Game Awards. The studio will likely emphasize the hand-crafted nature of the final assets to contrast with the AI placeholder work.
- Industry policy developments (2027): The International Game Developers Association (IGDA) and SAG-AFTRA are both pushing for clearer guidelines on AI use in game development. Crystal Dynamics' "pre-production only" approach could become a template that other studios adopt to avoid backlash.
- Competitor responses (2026–2027): Watch for statements from Naughty Dog, Rockstar Games, and CD Projekt Red — studios that have historically prided themselves on hand-crafted art. If they also quietly adopt AI for pre-production, the industry norm will shift decisively.
- Player reception at launch (2027): The ultimate test will be whether players can detect any AI "ghost" in the final game. If the hand-authored assets are indistinguishable from traditional pipelines, the controversy may fade. If not, expect a replay of the Foamstars backlash.
The Bigger Picture
This story sits at the intersection of two major trends: AI-assisted game development and the "AAA crunch" debate. The pre-production phase is historically one of the most wasteful in game development — teams build hundreds of placeholder assets that are thrown away, consuming time and labor that could go into final content. If AI can reduce that waste without displacing artists, it addresses a genuine pain point. But the line between "assist" and "replace" is thin, and studios that cross it face immediate reputational consequences.
The second trend is the normalization of AI as a collaboration tool rather than a replacement. Unlike the 2023–2024 wave of AI-generated game art that sparked union protests and player boycotts, Crystal Dynamics' approach reframes the technology as a productivity enhancer for human teams — similar to how procedural generation was adopted in the 2010s. The key difference is that procedural generation was always deterministic and controllable; generative AI is probabilistic and unpredictable. Whether studios can maintain that control at scale remains an open question.
Key Takeaways
- [Pre-Production Only]: Crystal Dynamics used generative AI exclusively for placeholder assets during pre-production, not for final game content — a critical distinction that may set a precedent.
- [Speed vs. Reputation]: The studio is trading on faster iteration cycles while explicitly distancing itself from the AI-generated final-content controversies that have hit Ubisoft and Activision.
- [Cross-Studio Alignment]: The tooling may have been particularly valuable for coordinating visual direction between Crystal Dynamics and Flying Wild Hog, two studios with different art styles.
- [Industry Template]: This "pre-production AI" model could become the standard approach for AAA studios seeking productivity gains without incurring player backlash — if the final product is indistinguishable from traditional pipelines.
