AI Sheriff's anniversary: Douyin's Large Model slashes rumor view count by 62%
On May 25, Douyin published its annual report on rumor governance, highlighting how AI large model technology has advanced within the information ecosystem. Since launching the "AI Quzhen Large Model" a year ago, the platform has achieved a major shift in fighting misinformation—moving from manual rumor detection to proactive model-driven action.

Data indicates that over the past year, the average view count of rumors identified and processed by the AI Quzhen Large Model decreased by 62%. This significant reduction means the window during which rumors can spread before being intercepted has been greatly shortened, effectively breaking the information chain early.
Technology Empowerment: Moving from Semantic Verification to Proactive Defense
According to the report, Douyin's AI Quzhen Large Model now features DeepResearch and multi-dimensional cross-verification. For example, when dealing with repetitive rumors that have multiple variations—like the false claim about monkeys being shot on Mount Emei—the model uses semantic recognition to automatically link and handle them, significantly easing the workload for human reviewers.

The key technical advances include:
Proactive Governance: Rumor identification accuracy has risen to 95%, and the response time for hot-topic events this year has improved by over 50% compared to previous levels.
Dual Protection: The large model not only retrieves authoritative sources but also deeply incorporates Douyin's years of manual review experience, ensuring precise judgment in complex logical reasoning.
Cognitive Defense: Using the "Quzhen Card" feature, the platform displays verified information from authoritative bodies and mainstream media as visual cards beneath questionable videos or within search results, directing users toward factual truth.
Value Return: Multi-party Collaboration for a Co-governance Ecosystem
A Douyin misinformation governance official said that the AI technology is designed to create a "cognitive defense" for users, but the ultimate foundation of rumor control remains news authenticity. The large model's effectiveness relies on timely and accurate clarifications from news subjects, official authorities, and media outlets.
The report also reviewed the top ten debunked rumors from the past year, covering public safety, health science, and social hot topics. Each case was traced and verified through AI cooperation with authoritative sources. Now, when users view controversial content or search related keywords, the system automatically triggers the "AI Quzhen" prompt, offering a full fact-checking source link.
Industry experts noted that Douyin's upgraded governance strategy not only showcases the engineering application of large models in content security but also signals a shift for short video platforms from simple "content blocking" to a "collaborative governance" model that fosters user discernment and jointly builds a healthier community.
Related article
Save up to $200 in 4 Days: The Second Reason to Attend TechCrunch Disrupt 2026
Time is ticking: you have just four days to secure up to $200 in savings on your TechCrunch Disrupt 2026 pass before prices rise on September 25 at 11:59 p.m. PT. Claim your ticket savings and enjoy 50% off a second pass here.This is the second compe
Soy MiniMax3 Monitor Speakers: Big Power in a Small Package
Sonic has launched the MiniMax3 Monitor, a new pair of professional studio speakers designed for desktop near-field monitoring and personal audio/video setups. Priced at 2099 yuan, these compact yet powerful speakers have quickly gained traction amon
Anthropic Launches Claude Mythos to Solve Erdős Conjecture With Concise Proof
On May 26, Sholto Douglas, an engineer at Anthropic, took to X to announce that their latest model, Claude Mythos, had cracked a century-old puzzle in combinatorial geometry: the Erdős unit distance conjecture. Dubbed a "clever and concise proof," th
Related Special Topic Recommendations
Comments (0)
0/500
On May 25, Douyin published its annual report on rumor governance, highlighting how AI large model technology has advanced within the information ecosystem. Since launching the "AI Quzhen Large Model" a year ago, the platform has achieved a major shift in fighting misinformation—moving from manual rumor detection to proactive model-driven action.

Data indicates that over the past year, the average view count of rumors identified and processed by the AI Quzhen Large Model decreased by 62%. This significant reduction means the window during which rumors can spread before being intercepted has been greatly shortened, effectively breaking the information chain early.
Technology Empowerment: Moving from Semantic Verification to Proactive Defense
According to the report, Douyin's AI Quzhen Large Model now features DeepResearch and multi-dimensional cross-verification. For example, when dealing with repetitive rumors that have multiple variations—like the false claim about monkeys being shot on Mount Emei—the model uses semantic recognition to automatically link and handle them, significantly easing the workload for human reviewers.

The key technical advances include:
Proactive Governance: Rumor identification accuracy has risen to 95%, and the response time for hot-topic events this year has improved by over 50% compared to previous levels.
Dual Protection: The large model not only retrieves authoritative sources but also deeply incorporates Douyin's years of manual review experience, ensuring precise judgment in complex logical reasoning.
Cognitive Defense: Using the "Quzhen Card" feature, the platform displays verified information from authoritative bodies and mainstream media as visual cards beneath questionable videos or within search results, directing users toward factual truth.
Value Return: Multi-party Collaboration for a Co-governance Ecosystem
A Douyin misinformation governance official said that the AI technology is designed to create a "cognitive defense" for users, but the ultimate foundation of rumor control remains news authenticity. The large model's effectiveness relies on timely and accurate clarifications from news subjects, official authorities, and media outlets.
The report also reviewed the top ten debunked rumors from the past year, covering public safety, health science, and social hot topics. Each case was traced and verified through AI cooperation with authoritative sources. Now, when users view controversial content or search related keywords, the system automatically triggers the "AI Quzhen" prompt, offering a full fact-checking source link.
Industry experts noted that Douyin's upgraded governance strategy not only showcases the engineering application of large models in content security but also signals a shift for short video platforms from simple "content blocking" to a "collaborative governance" model that fosters user discernment and jointly builds a healthier community.
Save up to $200 in 4 Days: The Second Reason to Attend TechCrunch Disrupt 2026
Time is ticking: you have just four days to secure up to $200 in savings on your TechCrunch Disrupt 2026 pass before prices rise on September 25 at 11:59 p.m. PT. Claim your ticket savings and enjoy 50% off a second pass here.This is the second compe
Soy MiniMax3 Monitor Speakers: Big Power in a Small Package
Sonic has launched the MiniMax3 Monitor, a new pair of professional studio speakers designed for desktop near-field monitoring and personal audio/video setups. Priced at 2099 yuan, these compact yet powerful speakers have quickly gained traction amon
Anthropic Launches Claude Mythos to Solve Erdős Conjecture With Concise Proof
On May 26, Sholto Douglas, an engineer at Anthropic, took to X to announce that their latest model, Claude Mythos, had cracked a century-old puzzle in combinatorial geometry: the Erdős unit distance conjecture. Dubbed a "clever and concise proof," th





Home






