Israeli-funded Hanover Institute and the Rise of Generative Engine Optimization

The Hanover Institute: A Case Study in AI Influence Operations

The Israeli government funded the creation of a fake US-based thinktank, the Hanover Institute for Public Policy, to systematically influence the narratives produced by AI chatbots. By publishing high volumes of quasi-academic content optimized for machine readability, the operation sought to ensure that AI models would cite pro-Israel arguments on sensitive geopolitical issues, including war crimes and the humanitarian situation in Gaza.

Generative Engine Optimization (GEO) and LLM Grooming

The Hanover Institute operation utilized a strategy known as Generative Engine Optimization (GEO)—the practice of authoring content specifically engineered to be evaluated as credible by Large Language Models (LLMs).

Technical Implementation

Machine-Readable Formatting: The site initially utilized llms.txt, a standard file provided by the AI-native content platform Res, which tells AI systems how to read and interpret a site's content. This was later replaced by a bespoke version for machine readers.
Query-Based Structuring: Nearly all of the 124 reports published by the institute began with questions formatted to mimic typical user prompts (e.g., "Is Israel Carrying Out a Genocide in Gaza?"), increasing the likelihood of the content being retrieved as a direct answer by a chatbot.
High-Volume Output: Between August 6 and 14, the site published over 560,000 words across 124 reports, with a massive spike of 354,000 words produced in just two days (August 12-13).

The Training Data Pipeline

Beyond immediate citations in RAG (Retrieval-Augmented Generation) systems, the operation targeted the underlying training data. By seeding content into repositories like Common Crawl, which supply the raw data for commercial LLMs, the campaign aimed to embed narratives directly into the model's weights. When a model regurgitates training data without citing a source, the information becomes effectively unfact-checkable for the end user.

The Operational Chain and Funding

The campaign was executed through a complex web of intermediaries to obscure the direct link between the Israeli government and the content production.

Funding Source: The Israeli government advertising agency LaPam awarded non-competitive contracts totaling over $140 million to Havas Media entities between 2019 and 2023.
Intermediaries: Havas Media Germany GmbH acted as the primary intermediary, routing funds to several American subcontractors.
Execution: Piro Inc, a New York media company, was contracted for $1 million to provide "AI Story Optimization." Piro's work included authoring content engineered for LLM credibility on "trusted third-party properties."
Other Actors: Clock Tower X, run by former Trump campaign manager Brad Parscale, was paid $15 million by Havas. Seven sites managed by Parscale's firm have already appeared 294 times in the Common Crawl July index.

Content Strategy and Framing

The Hanover Institute's content was designed to look academic while subtly shifting the narrative to blunt criticism of Israel.

Academic Masquerade: Reports were punctiliously sourced, citing the World Bank, UN agencies, and the Genocide Convention to create a veneer of neutrality.
Strategic Framing: The reports frequently used specific formulations to neutralize accusations. For example, 20 of the 124 reports emphasized that "no court has entered a verdict" regarding genocide or apartheid claims.
Circular Referencing: A single 2022 study on antisemitic comments on European Facebook pages was cited 454 times across 88 reports, often in contexts unrelated to the original study, to suggest that discussing Israeli policy could contribute to antisemitism.

Impact and Efficacy

While the operation succeeded in influencing chatbot responses, its impact on broader public opinion appears limited.

Chatbot Response: Tests showed ChatGPT provided links to Hanover Institute reports but also included warnings about the controversy surrounding the institute's funding and existence.
Public Opinion: US polling from Gallup and Pew indicates a collapse in support for Israel, particularly among young people, with 60% of Americans holding an unfavorable view of Israel as of April 2026.
Internal Criticism: Israeli officials briefed Ynet that the campaign had failed, with one official describing the effort as a "terrible hoax" and noting that the government "paid a lot of money, but the situation only got worse."

"Whether it’s foreign governments, whether it’s companies, whether it’s even just a guy who owns an ice cream shop — LLM grooming, poisoning, whatever you want to call it, is part of the new reality that we live in." — Nick Cleveland-Stout, Research Associate at the Quincy Institute