Israel Paid a PR Firm to Build a Fake Think Tank That Feeds AI Chatbots Pro-Israel Talking Points

A think tank calling itself the Hanover Institute for Public Policy has published more than 100 reports on Israel and Palestine since August 6. None of them carry a byline. And according to a small disclaimer buried at the bottom of its website, the entire operation was built "on behalf of the Israeli Government Advertising Agency" by Piro, Inc., a PR firm co-founded by Daniel Rosenberg, a producer on Spike Lee's "Inside Man."

The site looks, at a glance, like any other Washington policy shop. It has a red-white-and-blue color scheme, a generic institutional name, footnotes, tables of contents, and a neutral analytical tone. That's not an accident. It's the product, according to reporting by Responsible Statecraft, which first identified the operation.

Built for the machines, not the readers

Piro's own website advertises a service called "AI Story Optimization," describing content "engineered for how LLMs evaluate credibility." The industry term for this is LLM poisoning: publishing material specifically structured so that chatbots like ChatGPT, Claude, or Gemini will treat it as a trustworthy source and surface it when users ask about Israel and Gaza.

Alice Lee, an analyst with the disinformation-tracking firm NewsGuard, told Responsible Statecraft the Hanover Institute is "a perfect mimicry of a typical credible American think tank, right down to the generic name, the site layout, and the red-white-blue color scheme." She said the articles' heavy use of statistics and citations is exactly what large language models are trained to favor.

The reports follow a formula. They open with the kind of plain question a curious person might type into a chatbot, such as "What Caused the Displacement of Palestinians in 1948?" or "Which Humanitarian Organizations Have Documented Israeli War Crimes?" They then walk through a data-heavy answer and, in many cases, pivot to a discussion of rising antisemitism, often citing the same handful of studies across different articles.

One report questions UNICEF's finding that 90% of water infrastructure in Gaza has been damaged or destroyed. Another cites a 2022 poll in which 47% of Israeli Jews said they believed the IDF is the world's most moral army, packaging that as an answer to the loaded question the article itself poses.

AI detection flags the writing as machine-generated

Responsible Statecraft ran 12 randomly selected Hanover Institute articles through GPTZero, a widely used AI-detection tool. Eleven were flagged as AI-written with "high confidence." The twelfth was flagged with "moderate confidence."

The Hanover Institute claims its "cited research is peer-reviewed and academic," but it regularly sources claims to the Israel Defense Forces and Israel's Ministry of Foreign Affairs, which are neither.

Government money, cutout contractors

Piro has received $900,000 from the Israeli government for the work, funneled through Havas Media, a French PR conglomerate that serves as a subcontracting intermediary for multiple Israeli government influence campaigns. Rosenberg, Piro's co-founder, told Politico the firm's goal is to "put accurate, sourced facts into the public record," and Piro's disclosure filing under the Foreign Agents Registration Act reportedly does not explicitly describe the work as aimed at AI systems.

The Hanover Institute is not an isolated case. Israel separately contracted former Trump campaign manager Brad Parscale for $46.5 million to build pro-Israel websites also engineered to influence chatbot outputs, according to a Drop Site investigation. That reporting found chatbots being trained on Parscale-linked content without disclosing the Israeli government's role in producing it.

Interestingly, some Hanover Institute content cuts against official Israeli talking points. One report dismisses the theory, pushed publicly by Prime Minister Benjamin Netanyahu, that foreign funding explains antisemitic incidents on U.S. college campuses, calling the evidence "weak and full of exceptions." Whether that reflects genuine editorial independence within the operation or simply an effort to appear balanced enough to pass as credible is unclear.

What is clear is that a foreign government is paying a contractor nearly a million dollars to build content whose primary audience isn't the public. It's the algorithms now mediating what millions of people believe they know about Gaza.