China's ad industry issues first GEO standard after CCTV 'AI poisoning' exposé
China's advertising industry has published a five-part voluntary standard for generative engine optimization, six months after a CCTV consumer rights broadcast showed how easily AI assistants could be steered towards a product that did not exist. The March 2026 segment, aired on the annual 3·15 gala, featured an undercover insider who invented a smart wristband and seeded roughly two dozen promotional articles, reviews and rankings; at least two AI models later recommended the fictional device.
The series was issued by the China Association of Commercial Advertising through its AI Marketing Application Working Committee. It defines what GEO is, how results should be measured, what a GEO contract may promise, and where optimization ends and manipulation begins. The standards are marked T/, meaning they are voluntary industry agreements rather than law, and gain force when brands write them into contracts or regulators cite them. Drafters include Mininglamp, Zhihu, Sohu, Weibo's operator and 360, with Publicis Group named on parts one and five.
The measurement rules are the most portable. Because generative engines rarely return the same answer twice, the standard sets a floor of 30 runs at the finest level a vendor reports on, and warns that visibility scores built on five runs can swing 20 points with no real change. It also requires visibility to be calculated only on category questions, never branded ones, and demands disclosure of the model, prompts and rules behind any AI sentiment score.
Part five bans vendors from promising outcomes beyond technical reality, naming "absolute control of model output" and "100% screen dominance" as examples. For brands and agencies outside China, the practical takeaway is to ask any AI visibility supplier how many runs sit behind each figure and whether branded prompts are counted.
Reported by Search Engine Journal ↗
Topics: Measurement, Search policy, Citations