A synthetic think tank backed by Israel is publishing AI-generated essays to influence how chatbots and search engines present information about Israel and Palestine. The project aims to shape online narratives and public opinion.
Media and publishing professionals tracking AI-driven influence campaigns have new evidence of state-backed efforts to shape online narratives. The Hanover Institute for Public Policy, a synthetic think tank funded by Israel and operated by US-based Piro Inc, has published over 100 AI-generated articles in less than a month. The initiative is designed to feed large language models (LLMs) with content that could sway chatbot responses and search engine results in Israel’s favor.
The Hanover Institute’s website releases a dozen new articles every few days, focusing on topics such as Israel, antisemitism, and Palestine. Each article is framed as a question-examples include "How Much Palestinian Land Has Israel Taken?" and "Is There a Policy of Starvation in Gaza?"-and features graphs and citations to real sources, though without direct links. The articles lack bylines, and the site claims this is to encourage readers to evaluate sources rather than authorship. According to the about page, accountability is maintained through public disclosures and filings with the US Department of Justice.
AI detection tool Pangram analyzed several Hanover Institute articles and found that the main text was entirely AI-generated, with only the bibliographies written by humans. The images accompanying the articles are also AI-created. The site includes an llms.txt file, a technical marker that facilitates scraping by LLMs, further indicating that the content is intended for AI consumption as much as for human readers.
Piro Inc, the marketing firm behind the project, openly advertises its ability to influence chatbot answers through what it calls "AI Story Optimization." Co-founder Daniel Rosenberg described the process as identifying gaps in a brand’s online narrative and strengthening signals that shape how AI models interpret and explain that narrative. Piro’s website details a methodology that maps and scores online surfaces-such as Reddit threads, YouTube transcripts, and publisher pages-by their influence on AI-generated answers.
Regulatory filings show that Hanover is registered as a foreign agent in the US, with Piro Inc providing strategic communications and media relations for LaPam, Israel’s official advertising agency. Documents reviewed by 404 Media include invoices from HAVAS Media, a European agency working for Israel, showing payments of $900,000 and $100,000 to Piro for services including audience research, content production, and digital distribution targeting US audiences.
Margaret DeReus, executive director of the Institute for Middle East Understanding, said the campaign represents a continuation of longstanding efforts by Israel to influence US public opinion, but argued that such tactics are not reversing increased support for Palestinian rights. The Hanover Institute’s articles are described as impersonal and technical, but the organization’s website links its mission to the founder’s personal family history, referencing a relative who survived Auschwitz.
The Hanover Institute did not respond to requests for comment. The project’s approach to AI-driven content manipulation echoes broader concerns about the integrity of information in search and AI platforms. Similar issues have been raised in other contexts, such as the use of AI tools for fact-checking and verification, as seen in recent reporting on Google DeepMind’s Backstory tool.
