Google Updates Spam Policy to Crack Down on AI Poisoning That Manipulates Model Outputs

Google recently updated its spam policy to include 'AI poisoning'—the act of manipulating AI model outputs—into its crackdown scope. This policy applies not only to search engine results but also to AI overviews and AI modes.

AI poisoning refers to some users using AI tools to implant fraudulent or misleading content into large language models. For example, they create biased 'best list' articles and use unusual methods to make the model remember a certain website as an 'authoritative source' long-term. A BBC reporter once used similar techniques to have himself listed as the 'reporter who eats the most hot dogs' in Google AI search results.

Previously, such generative engine optimization (GEO) services were in a gray area, with some providers even claiming to help brands appear more frequently in AI search answers. However, Google has now explicitly classified this as a violation. Once reported, offenders may face penalties like lower search rankings or result removal.