A health tracker that by no means existed managed to land a spot on AI-generated “prime picks” lists in China, outranking actual merchandise with out ever being constructed.
The phantom gadget, referred to as Apollo-9, didn’t go viral accidentally. It was engineered into existence by means of a coordinated flood of faux critiques, rankings, and knowledgeable content material designed to sport AI techniques. Inside hours, chatbots started confidently recommending it.
The episode, uncovered in a state media investigation, provides a placing glimpse into how simply AI outputs could be formed and manipulated. It additionally places a highlight on generative engine optimization (GEO), a quickly rising tactic that goals to not rank on serps, however to quietly rewrite what AI believes is true.
How pretend information manipulated chatbot outcomes
An undercover investigation, aired throughout China Central Tv’s annual shopper rights gala on March 15, examined how fabricated on-line content material can form AI-generated responses.
Based on the South China Morning Submit, the report uncovered a system referred to as Liqing that was used to mechanically publish giant volumes of knowledgeable critiques, trade rankings, and person suggestions for a non-existent health tracker referred to as the Apollo-9.
After the content material was revealed, two AI chatbots have been requested to advocate good well being bracelets. Each chatbots talked about the fictional Apollo-9 and ranked it among the many prime selections, based on the published, although the report didn’t title them.
A hidden digital camera confirmed the operator behind the Liqing system boasting in regards to the observe. “We within the GEO trade are principally poisoning [AI],” he mentioned.
China Every day Hong Kong famous that findings from China Media Group additionally confirmed the identical technique practiced by a web based advertising instrument referred to as generative engine optimization, or GEO.
“Corporations have been utilizing GEO to mass-feed promotional articles in an try to govern the knowledge AI retrieves when answering person queries,” the report acknowledged.
GEO practices draw scrutiny as market grows
The case highlights the emergence of generative engine optimization, or GEO, which goals to affect how AI techniques retrieve and current data. The strategy mirrors search engine marketing however targets AI-generated solutions as a substitute of search rankings.
Analysis cited within the South China Morning Submit talked about that the GEO market reached about 250 million yuan, or about $36 million, in 2025 and is projected to develop quickly within the coming years.
The SCMP additionally highlighted that specialists mentioned the misuse displays early-stage trade habits. “That is really a typical drawback within the early levels of an trade,” mentioned Qian Wenying of the China Europe Worldwide Enterprise College.
Qian additionally defined that some corporations attempt to revenue by deceptive customers and flooding AI techniques with bulk spam content material.
China Every day Hong Kong reported that Li Fumin, a researcher at Shandong College of Finance and Economics, mentioned companies utilizing GEO on this approach are fabricating info and fascinating in misleading advertising.
“This type of habits permits individuals to obtain implanted advertising content material with out understanding it, which violates their shopper rights,” Li emphasised.
Regulation and safeguards beneath dialogue
The investigation triggered a sturdy public response in China, with associated hashtags trending on main social media platforms. Trade teams and regulators at the moment are weighing potential responses.
Based on the China Every day Hong Kong, specialists mentioned new guidelines may embody stricter oversight of content material assets, licensing necessities, and clearer accountability for platforms that host AI-generated or AI-influenced materials.
Some corporations have already moved towards self-regulation. Greater than 10 companies lately signed a GEO trade conference pledging to enhance content material high quality and keep away from manipulating algorithmic outcomes.
On the identical time, AI builders have responded cautiously. ByteDance mentioned its Doubao chatbot was not affected by such techniques, whereas Alibaba mentioned the reasoning capabilities of its Qwen mannequin weren’t impacted.
Learn extra about how AI brokers can leak information and expose governance gaps as enterprise adoption accelerates.
