Shenzhen’s market regulator has fined a small generative-engine optimization (GEO) vendor ¥50,000 for probing AI inclusion standards and fabricating rankings at scale. China’s March 15 gala had already spotlighted pay-to-appear “poisoning” of model answers; half a year later GEO did not vanish—it shifted from mass soft posts to preference probing. The case signals that even without classic false ads, polluting a model’s information environment and weakening neutrality can count as unfair competition against AI platforms.
From easy money to a black box
GEO bets that users trust confident, tidy AI answers, and that probabilistic models preferentially fit high-association, “credible-looking” material already in the corpus. Vendors pitch rewriting brand value into a knowledge base models will cite, promising compounding visibility unlike search ads that go to zero when spend stops.
Early playbooks were simpler than classic SEO: flood WeChat, Toutiao, Baijiahao, Xiaohongshu, Zhihu and similar with soft articles. A year or two ago Doubao, DeepSeek, Yuanbao and peers were hungry for external content, so more AI-written posts meant higher odds of appearing in answers. After platforms weighted authoritative sources, spam failed; vendors paid niche industry and local sites instead, then those sources got watery too and external inclusion rules went opaque. This summer many GEO shops collapsed when posts kept shipping but brands stopped being cited. For an earlier take on GEO’s marketing impact, see Is GEO wrecking AI marketing?.
Why “analysis systems” crossed the line
Unlike Baidu or Google’s public quality guidelines, most AI vendors do not publish how they ingest outside material. Some operators built “GEO analysis systems” that probe answer preferences and how products score authority, density and source trust, then sell those “inclusion standards.” The Shenzhen case targets exactly that probing plus fake rankings. Regulators’ logic: polluting the information environment and eroding neutrality can put a vendor in competitive conflict with the AI firm—hence unfair competition.
The piece’s bet: GEO can still move outputs, but operating costs keep rising; many shops may die from cost before from total ineffectiveness. Against slogans that “SEO is dead, GEO forever,” see the AI search era. Durable visibility still favors verifiable facts and clean site structure over gaming a hidden rulebook.
FAQ
Q: What was the fine and the charge? A: Shenzhen’s market regulator fined a small GEO vendor ¥50,000 for probing AI inclusion standards and fabricating rankings in bulk.
Q: How do vendors claim GEO differs from search ads? A: They say ad spend is rental traffic that vanishes when it stops, while GEO tries to encode the brand as knowledge models will reuse. That pitch does not legalize probing inclusion rules or fake rankings.
Q: Why did soft-article flooding fade? A: After vendors raised the weight of authoritative sources, generic platforms and watered-down industry sites lost priority; external rules grew opaque, so pure volume got worse economics.
相关文章
