Fact Check: AI mass production of anti-China fake videos by Japan

Editor's Note: In the past two years, a large number of orders for producing anti-China fake videos have emerged on the Japanese crowdsourcing platform CrowdWorks. Relevant parties have used AI tools to mass-produce fabricated storylines and content, forming a commercialized rumor-mongering industrial chain. According to Asahi Shimbun's investigation, the creators involved admitted that all the content was entirely fabricated from scratch. Driven by profit incentives, such misleading information continues to spread. The following two typical videos serve as examples, clearly illustrating the visual characteristics of this "AI mass production" model in specific content.

Claim

In March 2025, YouTube users @Japanのミカタ【海外の反応】 and @ニッポン応援エピソード【海外の反応】uploaded videos on the platform. The core theme of both videos was to accuse Chinese tourists of damaging cherry blossoms during Japan's cherry blossom season. The footage in both videos consists of multi-source mixed editing (collage of various clips).

Fact Check:

1. Visual Analysis

1) Visual analysis of an image extracted from the video posted by @ニッポン応援エピソード【海外の反応】 reveals multiple telltale signs of AI-generated synthesis:

Figures: Deformed fingers, abnormal finger counts and distorted facial features; asymmetrical eyes and a warped nose; skin that is overly smooth and devoid of realistic shading.

Overall scene: An excessively crowded composition with exaggerated, repetitive poses; blurry, patchy cherry blossom edges; and unnatural falling petal trajectories.

2) In the video posted by user @Japanのミカタ【海外の反応】, images bearing traces of AI generation were also discovered. One of these images is analyzed as follows:

Cherry blossoms: Branches exhibit bilateral symmetry and repetitive textures; distant blooms are blurry and lack definition.

Figures: In medium-to-long distance shots, head proportions are abnormal and faces dissolve into color blobs; shoulders and arms appear to "melt" into one another with indistinct boundaries; the crowd is excessively dense and unnaturally structured.

In addition to the visual AI artifacts, both videos also exhibit clear signs of manual post-editing and splicing in their narrative pacing and shot transitions.

2. AI detection results

The AI detection results for the two images using Hive Moderation are as follows:

- Image from @Japanのミカタ【海外の反応】: 99.9% likely to be AI-generated 

- Image from @ニッポン応援エピソード【海外の反応】: 98.4% likely to be AI-generated

Combining visual analysis with technical detection results, the way these two videos were produced closely matches the systematic fabrication method uncovered by the Asahi Shimbun's investigation – specifically, a process where "the client mandated the use of AI image generation tools."

3. Reports from The Asahi Shimbun

The Asahi Shimbun newspaper published two in-depth reports on Japan's "anti-China" video industry. Through interviews with various creators – including a former national civil servant in his 60s and young part-time workers – the creators admitted that all the videos were completely fictional, with no on-site filming or fact-checking.

Read more: Click-driven hate: AI-generated anti-China videos form profit chain in Japan