Level-5 AI Trailer Backlash: What the Studio Admitted, What It Promised, and Lessons for Creative Teams
After fans flagged suspected AI imagery in its VISION 2026 II showcase, Level-5 apologised, explained which flaws were human error and pledged to avoid direct AI use in videos for fans. A summary of the statement, the contrasting Higgsfield Adathon win, and practical disclosure lessons.
In September 2026, Japanese studio Level-5, known for Professor Layton and Yo-kai Watch, faced a strong fan backlash over suspected generative AI imagery in its showcase trailers. Within a week it published a formal statement apologising, explaining which apparent flaws were human error, and promising to stop using AI directly in videos delivered to fans. In the same week, an advertising competition run by AI video platform Higgsfield awarded its top prize to a spot that jokes about clients no longer wanting AI. Together, the two stories show how differently audiences treat AI in games marketing and in advertising, and why clear disclosure matters. This article summarises what happened, what Level-5 actually said, and the lessons for creative teams.
Timeline of the Level-5 backlash
According to IGN, events unfolded as follows:
| Date (2026) | Event |
|---|---|
| September 9 | A Nintendo Direct includes trailers for Yo-kai Watch 2: Haunted Domain and a Switch remake of Professor Layton and the Curious Village. |
| September 10 | Level-5 releases its “VISION 2026 II” showcase on YouTube. Viewers flag inconsistent backgrounds and details they attribute to generative AI, and compare the Yo-kai Watch trailer with the Nintendo Direct version from the day before. |
| Following days | CEO Akihiro Hino posts on X that the studio experimented with AI processing for the showcase out of a desire to make the presentation flashy, while saying story, character designs and settings were made by people. The response is largely negative. |
| September 17 | Level-5 publishes a formal statement on its website apologising and explaining specific issues. |
IGN reports that Hino also said AI tools were helping the company try to shorten development of major titles from about five years towards two. Many commenters, in Japanese and English, objected that a trailer should express creativity rather than flashiness, and raised concerns about how generative models are trained on artists’ work.
What Level-5’s statement says
The September 17 statement, published in Japanese, makes several specific points:
- Re-uploaded trailers. Some trailers shown in VISION 2026 II would be re-uploaded at the start of Tokyo Game Show in corrected versions, fixing both errors caused by AI use and human mistakes that looked like AI errors.
- Human error mistaken for AI. For a Yo-kai Watch 2 key illustration, the studio said power lines criticised as unnatural were hand-drawn by its designer and intentionally simplified; it showed excerpts of production materials and redrew them to look more natural. It said details on a vending machine and rain gutters were similar simplification errors, and that bicycle brake levers were also missing in the earlier Yo-kai Watch 2 game, though it should have drawn them this time.
- Character model changes. Fans noticed that protagonist Keita’s face differed between the Nintendo Direct and Level-5’s own showcase. The studio said no AI was used for this; the model was revised for quality, but the Nintendo Direct deadline came first, so the new version appeared only in its own presentation.
- Scope of AI use. It stated there is no direct use of AI in game content, and that AI was not used for Professor Layton and the New World of Steam.
- Apology. It apologised for causing distrust through immature use of AI and simple human mistakes.
- Future policy. It promised to keep AI use within the scope of efficiency improvements and to avoid direct use in videos delivered to users.
Nintendo Life noted that the statement confirmed New World of Steam contains no AI but gave no equivalent confirmation for the Curious Village remake, where fans had raised concerns that backgrounds looked like AI upscales of the originals.
The contrast: an AI ad that jokes about AI
On September 16, ADWEEK reported that AI-R Studio, the AI and hybrid production studio of Mexico-based Pico.Love OS, had won the inaugural Adathon, a month-long competition hosted by Higgsfield in partnership with ADWEEK and announced at Brandweek 2026 in Atlanta. Teams wrote a brief and had four weeks to produce a spot using AI. The winning entry opens with AI-generated footage of a toucan and a gorilla in a jungle before a director shouts, “Cut! I’m so sorry guys, they no longer want AI.”
The two stories point in opposite directions. In advertising, AI production is openly celebrated and even self-aware about client resistance. In games, where fans have strong attachments to hand-crafted art styles, the suggestion of undisclosed AI use was enough to damage trust in a long-established studio.
Why the backlash was so strong
Several factors combined:
- Disclosure came after detection. Fans found the anomalies first; the explanation followed. That ordering makes any later clarification look defensive.
- Mixed causes. Some issues were attributed to AI and others to human error. Once audiences suspect AI, ordinary simplifications in hand-drawn art are also read as AI artefacts, which Level-5 itself acknowledged.
- Franchise expectations. Long-running series carry expectations about art style and craft. Remakes that alter original backgrounds face extra scrutiny.
- Framing. A stated motive of making the presentation flashy reinforced the impression that AI was used for surface polish rather than creative purpose.
Lessons for creative teams
Whether a team is for or against AI, the Level-5 episode suggests practical steps:
- Decide and document AI use before release. Record which assets used AI tools and how, so questions can be answered quickly and accurately.
- Disclose proactively for audience-facing work. A short note on how AI was used is easier than an apology after fans investigate.
- Keep production records. Level-5 was able to show production materials to explain some issues. Version histories, layered files and sketches are useful evidence; we discuss this in our guide to documenting your creative process.
- Review for artefacts. Whatever the tool, check final outputs for inconsistencies that audiences will notice.
- Consider provenance tools. Content credentials and watermarks can support claims about how work was made. See our articles on AI art authentication and provenance tracking and EU AI labelling rules and SynthID.
- Match policy to audience. What works in advertising may not work for a fan community. Level-5’s new policy draws a line at direct AI use in videos delivered to users while allowing efficiency uses behind the scenes.
For more on how AI imagery is being sold and displayed, see our pieces on the Firefly and Nano Banana image marketplace and AI art galleries and exhibitions.
Our analysis
Level-5’s statement is notable for its specificity: it named individual flaws, separated AI errors from human ones, and set a clear policy for future fan-facing videos. That is more than many companies offer. The episode still shows that trust is easier to keep than to rebuild. Studios that use AI in any part of their pipeline should assume audiences will look closely, and should be ready to explain their process before being asked. The Adathon result shows that AI-made work can be well received when it is open about its methods; the problems arise when audiences feel they discovered something that was not disclosed.