Israel takes its war on Palestine into the AI age

The battle over Palestine has never been fought only on the ground. For decades, Israel and its powerful political, lobbying and public-relations networks have devoted enormous resources to shaping how the conflict is understood in Washington and across the Western world. But an investigation into an Israeli-funded influence operation suggests that this campaign has entered a new and potentially more consequential arena: the artificial intelligence systems increasingly used by millions of people to determine what is true.

Recent investigations exposed the Hanover Institute for Public Policy as a purported American think tank that appeared to have no real offices, identifiable scholars or meaningful institutional presence. Instead, the institute appears to have been created as a sophisticated propaganda operation aimed at shaping search engine results and influencing the answers generated by AI chatbots.

Documents filed under the Foreign Agents Registration Act (FARA) trace the operation’s funding through advertising and public-relations firms to LaPam, Israel’s state advertising agency. Piro Inc, the New York advertising firm involved, was operating under a $900,000 digital communications work order on the Israeli account.

What makes the case extraordinary is not simply that an Israeli government-funded operation sought to influence public opinion. Governments have always done that. What is different is the apparent attempt to influence the machines increasingly shaping public opinion.

In just nine days, the Hanover website reportedly produced more than 560,000 words across 124 academic-style reports. The material was structured around questions likely to be submitted to AI systems, including whether Gaza faced a policy of starvation and whether the Israeli military is “the world’s most moral army”.

Piro markets this approach as “AI Story Optimization”, designed to make material highly visible to search engines and easily digestible to large language models. The site also included technical features designed to make its material easier for AI systems to find and process.

The objective, therefore, was not simply to persuade readers. It was to influence the algorithms that increasingly determine what information those readers receive.

Traditional propaganda can often be identified by its source, tone or obvious political agenda. More sophisticated propaganda is harder to detect because it disguises advocacy as neutral expertise. The Hanover material reportedly employed statistics, charts and graphs, cited legitimate primary documents and adopted the language of academic research while selectively framing evidence and omitting external links.