要点
- AI-generated fake stories (animals, etc.) are produced on a conveyor belt in content farms that look like real media.
- Programmatic ad systems finance this ecosystem by placing brands' ads on AI content farms, often without the brands' knowledge.
- Research shows ~50% of new English articles from 2025 show signs of machine generation; search engines may demote them but volume is rising.
- In Brazil, fake AI doctor videos sell harmful health products; Ukraine and other countries face similar risks.
数字で見る
Verification platforms like Snopes discovered that AI-generated fake stories about animals were being produced like a conveyor belt from a Facebook page. This content was published by fully automated "content farms" that looked like real news sites. In a NewsGuard experiment, a fully automated news site was set up in two days for $105 with the help of a Pakistani freelancer, and this model was quickly replicated.
The programmatic advertising infrastructure finances the ecosystem by automatically placing big brands' ads on these fake sites. In NewsGuard's two-month monitoring, ads from 141 major brands appeared on AI content farms. At the same time, while propaganda structures like Russia's "Pravda" network poison language models, commercial actors are also flooding the internet with garbage content produced solely for clicks and impressions.
Analyzing 65,000 English pages, Graphite found that from 2025 onward, about half of new articles bore signs of machine generation; however, this volume increase reflects only publication count, not actual read rates. In Brazil, the BBC uncovered more than 50 YouTube channels selling harmful health products by producing fake AI doctor videos. In Ukraine, although not yet widespread, the existing fake drug market is laying the groundwork for this model's spread.
リアクション
この記事について質問
回答はこの記事のみからAIが生成します。
よくある質問
- How do AI fake contents make money?
- Programmatic advertising infrastructure automatically places big brands' ads on AI content farms, generating revenue; brands usually don't know which sites their ads appear on.
- How widespread are these fake contents?
- Graphite analysis shows about half of new English articles from 2025 onward exhibit machine-generated characteristics; this reinforces the "dead internet" theory.
- Is there a danger for Turkey or other countries?
- Although the news is Ukraine-focused, the Brazil example shows fake AI doctor/expert videos are a global trend; local fake drug/cure markets could adopt this model.