Israel’s $900,000 U.S. pilot reached ChatGPT through Hanover articles

Israel’s Government Advertising Agency funded a U.S. digital-content pilot whose material later appeared in answers from ChatGPT and Perplexity during tests by Politico. Politico reported on August 14, 2026 that more than a dozen articles from the Hanover Institute were part of a $100,000 campaign, and that both AI tools cited them in neutral prompts.

The evidence shows a government-backed effort to distribute research-styled content into the U.S. information ecosystem, not proof that Israel altered either tool’s training data, rankings, or answers at scale. That distinction is the whole ballgame: being cited is evidence of visibility, not evidence of control.

$100,000 Hanover Content Campaign Sat Inside Piro’s $900,000 Pilot

A Piro, Inc. filing under the Foreign Agents Registration Act (FARA) identifies Havas Media Germany as acting for Israel’s Government Advertising Agency, also called LaPam, in its work with the U.S. firm. The April 30, 2026 work order authorized a $900,000 “Digital Storytelling Pilot” aimed at U.S. digital and social-media distribution.

The filing itemizes $190,000 for strategy and content development, $80,000 for production oversight, $400,000 for production hard costs, $95,000 for distribution and program evaluation, $35,000 for media consultation, and $100,000 for measurement and analytics. Those categories are disclosed in the FARA work order.

Diagram showing Israel's Government Advertising Agency working through Havas Media Germany and Piro to distribute Hanover Institute content

Politico’s reporting adds the missing link between that broad pilot and the Hanover Institute: it says more than a dozen Hanover articles were tied to a $100,000 campaign. The FARA work order itself does not name Hanover, ChatGPT, Perplexity, or AI-search citations. It describes a distribution project; Politico identified the specific material later tested in AI answers.

Hanover presents its work in the language of research rather than advocacy. A page asking whether antisemitism rises after conflict describes the institute’s work as nonpartisan and nonadvocacy, while carrying this campaign disclosure:

“This material is distributed by Piro, Inc. for Havas Media Germany on behalf of the Israel Government Advertising Agency (LaPam).”

Another Hanover article, “Is Anti-Zionism Antisemitism? What the Evidence Measures,” covers one of the subjects Politico said appeared in citations from the AI tools. The label tells readers who paid for distribution; it does not independently validate the articles’ research claims.

Conceptual illustration of an AI answer citing a research-style article carrying a sponsored distribution disclosure

Politico tested ChatGPT and Perplexity with neutral queries and found citations to the Hanover material on themes including antisemitism and anti-Zionism. That is a meaningful result because AI-answer visibility is an increasingly valuable distribution channel: Pew Research Center found in June 2025 that 34% of U.S. adults had used ChatGPT, roughly double the share reported in 2023.

Still, one set of citations does not show that the campaign changed what either system learned or reliably surfaces. Large language models may draw on indexed web pages, retrieval systems, user-specific context, and changing search results; a citation can appear because a page is available and relevant to one prompt. Politico’s tests establish that the content could get into answers, not that it shifted a platform’s underlying information environment.

The primary filing is Piro’s sworn account of its work for the foreign principal, while the AI-citation test comes from Politico’s independent reporting. Neither source establishes that Hanover content had a systematic effect across users or prompts.

What the disclosures do show is more concrete: a foreign government’s advertising arm paid for U.S. digital storytelling through an agency chain, and some of that disclosed material subsequently turned up in leading AI assistants. For a campaign built to shape what people encounter online, that is not a hypothetical distribution route, it is one it demonstrably reached.

Key Takeaways

Israel’s Government Advertising Agency was identified in Piro’s FARA filing as the foreign principal behind a $900,000 U.S. Digital Storytelling Pilot.
Havas Media Germany acted between Israel’s advertising agency and the U.S. firm Piro.
Politico reported that more than a dozen Hanover Institute articles were part of a $100,000 campaign.
Politico found ChatGPT and Perplexity citing Hanover material in neutral tests.
Those citations do not prove changes to model training data, rankings, or AI answers at scale.