We Investigated Whether Clair Obscur: Expedition 33 Used AI—Here’s The Verdict

Quick catch-up on a touchy subject: yes, Clair Obscur: Expedition 33 used generative AI at some point during development. The studio says it was limited. The internet, predictably, is not buying simple answers.

What Sandfall actually said

Back in July 2025, Sandfall Interactive told EL PAÍS that AI-assisted tools were part of the production pipeline. COO and production director François Meurisse framed it as targeted, practical help that let the team keep costs in check without cutting corners.

"We use some AI, but not much. The key is that we were very clear about what we wanted to do and what to invest in. And, of course, technology has allowed us to do things that were unthinkable a short time ago."

He also said the studio kept Expedition 33’s budget under $10 million, which, if true, is kind of wild given how polished the final game looks. How central AI was to reaching that goal, though, still isn’t clear.

The awkward timing and the double standard

This all bubbled back up right as Larian Studios was getting hammered for its own AI use. Insider Gaming’s Tom Henderson pointed out the hypocrisy: the industry is roasting Larian while cheering Expedition 33, which had just snagged nine wins at The Game Awards, including Game of the Year. The timing could not have been worse for anyone trying to keep this debate calm.

The visible proof players caught

Sandfall hasn’t broken down exactly where or how AI showed up in production, but there is one on-the-record example. After launch, players found a couple of posters in the opening area that had the usual AI tells: warped lettering, weird composition, that uncanny sense of something being almost right but not actually right. Those assets were removed quickly in an early patch. The running theory is that they were placeholders that accidentally shipped in the 1.0 build. On April 30, 2025, one player even flagged that the posters had been swapped out for custom art, and that another similar image in the starting zone had been removed too.

So yes: there was AI somewhere in the workflow. And aside from those posters and the studio’s own comments, that’s the only concrete evidence players have.

Mixed messaging: creation vs tools

Here’s where it gets messy. Meurisse says the team used some AI, sparingly. But in a later interview cited by GamePressure, CEO and creative director Guillaume Broche gave a very different impression, saying the studio draws a firm line when it comes to anything that qualifies as creative work. In other words: using tools internally is one thing; generating final creative content is another. Depending on how you define creative, that leaves wiggle room big enough to drive a truck through.

How this actually unfolded

• April 30, 2025: Players report that a couple of AI-ish posters in the opening area get patched out and replaced with custom assets; another similar poster in the starting zone disappears too.
• July 2025: In an interview with EL PAÍS, Sandfall confirms AI-assisted tools were used during development. Meurisse says it was limited and helped keep the budget under $10 million.
• December 2025: Expedition 33 cleans up at The Game Awards with nine wins, including Game of the Year.
• December 17, 2025: As Larian takes heat for AI, Tom Henderson highlights the double standard, noting that the new GOTY also used generative AI.

The line everyone keeps trying to draw

Welcome to 2025 game development, where small teams try to build giant games without giant budgets. AI tools offer a tempting shortcut, and when the end result is as slick as Expedition 33, barking for a total ban can feel disconnected from reality. The flip side is just as real: normalize it now, and it gets a lot easier for bigger studios to replace artists later.

From the outside, Sandfall seems to have kept AI on the internal side of production and scrubbed it from final assets as soon as the mistake popped up. Whether that’s a responsible compromise or a slippery slope depends on your threshold for risk vs results.

Where do you land on this? Draw a hard line early, or let smaller teams use every tool they can to punch above their weight?