Attackers Poison AI Chatbots Using SEO-Driven Disinformation

Threat actors are manipulating artificial intelligence chatbots, including ChatGPT, Google Gemini, and Google AI Overviews, by seeding the web with malicious links and disinformation.

The campaign involves creating and optimising content on various websites to ensure it is indexed by search engines and subsequently ingested by large language models (LLMs). By using search engine optimisation (SEO) techniques, attackers can influence the answers provided by these AI systems.

The primary objective of this manipulation is to conduct large-scale phishing attacks and to distribute disinformation. When users query an AI for information or recommendations, the model may present attacker-controlled links as part of a factual response or a legitimate source.

This method exploits the way modern AI tools integrate real-time web data to provide updated answers. By poisoning the web-based data sources that AI agents crawl, attackers can bypass some traditional security scrutiny, as users may be more likely to trust a link embedded within a direct, synthesised AI response.