Israeli government funded a $46.5 million digital influence campaign aimed at shaping how major AI chatbots respond to questions about the war in Gaza. According to the report, former Trump campaign manager Brad Parscale and his firm, Clock Tower X, created a network of websites publishing hundreds of articles designed primarily to be indexed by search engines and surfaced by AI systems rather than read by human audiences. The report says the campaign targeted AI platforms including ChatGPT, Claude, Gemini, and Perplexity.
The investigation alleges that the strategy focused on what is sometimes called generative engine optimization (GEO)—creating large volumes of web content intended to influence the information retrieved or learned by AI models. Internal documents cited by the report reportedly describe goals such as delivering preferred "GPT framing results" in chatbot conversations. The articles promoted narratives defending Israeli military actions, questioned some widely reported incidents from the Gaza conflict, and emphasized stronger U.S.-Israel cooperation.
The report further claims that some AI systems have already referenced websites created under the campaign, although the extent of their influence remains unclear. Large language models rely on a mix of pretraining data, search indexes, retrieval systems, safety policies, and human feedback, making it difficult to determine how much impact any single website network has on responses. At the time of reporting, the investigation presented evidence of the campaign's existence and objectives but did not independently establish the overall effect across different AI platforms.
The investigation has renewed debate over transparency and information integrity in the AI era. Experts say governments, companies, advocacy groups, and commercial organizations are increasingly trying to influence what AI systems retrieve or generate by publishing optimized online content. The allegations underscore broader concerns about the need for AI developers to evaluate source quality, diversify information sources, and strengthen safeguards against coordinated attempts to manipulate AI-generated answers on politically sensitive topics.
