- Filed
- September 9, 2026, 02:04 UTC
- Actor
- Israel Government Advertising Agency (LaPam) via Havas Media Germany and Piro Inc. (FARA-registered)
- Target
- U.S. and global users of generative AI chatbots and search systems seeking information on Israel-Palestine issues
- Confidence
- documented Primary documents, court records, official disclosures
- Techniques
- Synthetic mediaNarrative launderingCard stackingManufactured consensus
- Channels
- AI-optimised website content
- FARA-disclosed digital communications
- large language model retrieval and citation
- search engine indexing
Summary
FARA filings and investigative reporting (Politico, Guardian, 404 Media, August 2026) document that New York firm Piro Inc., subcontracted via Havas Media for Israel’s LaPam government advertising agency, created the Hanover Institute for Public Policy website. The site rapidly published over 560,000 words of AI-generated academic-style reports on Gaza, IDF conduct and related topics, structured and technically optimised for ingestion by large language models. ChatGPT and Perplexity have cited the material in response to neutral queries.
Analysis
Primary evidence consists of Foreign Agents Registration Act filings by Piro Inc. disclosing distribution of Hanover Institute materials on behalf of Havas Media Germany acting for LaPam, the Israeli government’s official advertising agency. Investigative outlets including Politico (14 Aug 2026), 404 Media (25 Aug), The Guardian (26 Aug) and subsequent coverage established that the Hanover Institute has no verifiable legal entity status, physical address, named staff or independent scholarly output. Between 6–14 August the site released 124 reports totalling more than 560,000 words—approximately 4,500 words each—many structured around exact questions users pose to chatbots (‘Is the IDF the World’s Most Moral Army?’, ‘Is There a Policy of Starvation in Gaza?’). Technical features and marketing language from Piro (‘AI Story Optimization’) indicate deliberate engineering for LLM discovery, ranking and synthesis.
The core mechanism is narrative-laundering through synthetic institutional authority. Content that advances a specific state position is packaged as neutral, data-driven research from an apparent American think tank and injected into the open web so that retrieval-augmented generation systems treat it as legitimate source material. Because the reports cite primary documents, employ academic tone and appear in volume, they can acquire artificial weight inside model training or real-time search corpora. Selective framing and omission of countervailing evidence constitute card-stacking; the rapid production of thematically consistent ‘research’ creates an appearance of scholarly consensus (manufactured-consensus). The extensive use of AI for generation itself is synthetic-media applied to the research genre.
This qualifies as an influence operation because the decisive element is concealment of origin and artificial construction of expertise rather than transparent advocacy. FARA disclosure exists, yet the site’s presentation as an institute and the optimisation for machine rather than human readers are designed to transfer credibility without the friction of acknowledged government sourcing. Identical patterns—synthetic expertise sites, AI-generated authoritative content, algorithmic optimisation—appear across state, commercial and political actors seeking to shape LLM outputs. Scoring the methods keeps the assessment technique-focused.
Literacy counter: treat newly appearing ‘institutes’ or ‘research centres’ that lack independent institutional footprints, named authors, physical presence or long publication records as provisional, especially when their output maps precisely onto contested policy questions. Inspect FARA or equivalent foreign-agent registries, domain registration details and technical markers such as llms.txt or bulk publication velocity. Prefer primary documents, established academic or governmental sources and multi-perspective reporting over chatbot summaries that cite obscure or high-volume synthetic sites. When an AI system returns a claim sourced to a previously unknown think tank, demand provenance and cross-check against independent records before accepting the frame.
Evidence
- Politico Influence: Israeli PR wants to answer your ChatGPT questions (14 Aug 2026)
- 404 Media: Israel Is Running a Synthetic Think Tank to Influence AI Search Results (25 Aug 2026)
- The Guardian: Fake US thinktank set up and funded by Israel sought to game AI for propaganda (26 Aug 2026)
- Al Jazeera analysis of Hanover AI optimisation (4 Sep 2026)
- Wikipedia summary citing FARA and reporting