Forensic Research Archive of Manipulation and Exploitation
Illustration: a dossier collage of a typewriter, drafted press releases and an annotated speech transcript.

Watch briefs · September 11, 2026

Israel Government Advertising Agency funds synthetic Hanover Institute think-tank site optimized to seed pro-Israel framing into LLM retrieval and chatbot answers

Illustration, not an archive object

Filed
September 11, 2026, 02:00 UTC
Actor
Israel Government Advertising Agency (LaPam) via Havas Media Germany GmbH and Piro, Inc. (FARA-registered)
Target
US and English-language public opinion mediated by large language models, search engines and AI chatbots on Israel–Palestine issues
Source
Unidentified Source blurred: contractors, cutouts, buried disclosure
Confidence
documented Primary documents, court records, official disclosures
Techniques
Corpus poisoningNarrative launderingSynthetic mediaManufactured consensusDisclosure theater
Where
United States, Israel
Channels
  • Hanover Institute for Public Policy website (hanoverinstitute.org or equivalent)
  • FARA-registered contractor chain (Piro → Havas → LaPam)
  • AI Story Optimization / Res platform tooling for LLM citation
  • question-formatted reports designed as chatbot prompts
  • search-engine and Common-Crawl indexing pathways

Summary

FARA filings by New York firm Piro, Inc. and reporting by Politico, The Guardian, 404 Media and Responsible Statecraft document that the Hanover Institute for Public Policy—a website with no legal entity, no physical address, no named staff and no bylines—was created on behalf of Israel’s LaPam (Government Advertising Agency) via Havas Media Germany and Piro. Between 6–14 August 2026 the site published 124 academic-style reports totaling more than 560,000 words, structured as question-form prompts typical of chatbot queries on Gaza, starvation, IDF morality, antisemitism and related topics. Content was produced with AI assistance and engineered via ‘AI Story Optimization’ tools for citation by ChatGPT, Perplexity, Claude and Gemini. Tests by journalists confirmed chatbots cited the material. A bottom-of-page disclosure identifies the foreign-agent chain; the design goal is corpus poisoning of retrieval systems rather than direct human persuasion.

Analysis

Primary records are the FARA filings submitted by Piro, Inc. in August 2026 disclosing work ‘on behalf of Havas Media Germany GmbH on behalf of the Israel Government Advertising Agency (LaPam)’ for the creation and dissemination of informational materials, together with contemporaneous open-source investigations by Politico (14 Aug), 404 Media (25 Aug), The Guardian (26 Aug) and Responsible Statecraft. The Hanover site presents as a conventional American public-policy institute: neutral tone, footnotes, graphs, academic formatting, red-white-blue palette. It has no registered legal entity, no street address, no named scholars or staff, and every report lacks a byline. In nine days it emitted 124 reports (>560 k words), many titles phrased as the exact questions users put to chatbots (‘Is there a policy of starvation in Gaza?’, ‘Is the IDF the world’s most moral army?’). AI-detection tools flagged the bulk of the prose as machine-generated; only bibliographies showed heavier human editing. Piro’s own marketing describes ‘AI Story Optimization’ engineered for how LLMs evaluate credibility; the site was built on a commercial platform (Res) whose purpose is to improve citation probability by generative systems.

Mechanism: corpus-poisoning is the load-bearing technique—seeding indexed, citation-shaped text so downstream retrieval and generation systems treat the preferred frame as independent research. Narrative-laundering occurs when the state message travels through a performed U.S. think-tank surface rather than an Israeli government channel. Synthetic-media supplies the volume and polish that no human newsroom could sustain at that cadence. Manufactured-consensus is attempted by the sheer density of ‘reports’ that appear to converge on the same conclusions. Disclosure-theater is present in the form of a technically compliant FARA footnote and bottom-of-page statement that does not travel with the content once a chatbot or search snippet extracts a paragraph.

This is an influence operation and propaganda because origin is concealed behind institutional mimicry, the content is engineered for machine rather than human consumption, and the design goal is to shape foreign (primarily U.S./English-language) belief via intermediaries the audience trusts more than a labeled government press release. Technique scoring is actor-agnostic: identical LLM-optimized seeding, synthetic think-tank form and contractor-chain laundering appear in other state-linked campaigns already documented this cycle. Color is gray: FARA makes the ultimate sponsor visible to anyone who reads the filing, yet the ordinary chatbot user or search-result consumer encounters only the think-tank surface.

Literacy counter: when a sudden cluster of polished, question-titled, byline-free ‘reports’ from a previously unknown institute appears on contested topics and is rapidly cited by chatbots, inspect the legal entity, staff page, physical address and FARA or equivalent registration. Prefer primary documents, named scholars with verifiable affiliations, and independent replication over volume-optimized synthetic dumps. Demand that provenance tags travel with every extracted snippet. Apply the identical corpus-poisoning test to any faction’s LLM-optimized institutes—the method does not change with the flag.

Tells

Evidence

  1. Politico Influence: Israeli PR wants to answer your ChatGPT questions (14 Aug 2026)
  2. The Guardian: Fake US thinktank set up and funded by Israel sought to game AI for propaganda (26 Aug 2026)
  3. 404 Media: Israel Is Running a Synthetic Think Tank to Influence AI Search Results (25 Aug 2026)
  4. Responsible Statecraft: Israel creates fake think tank in likely attempt to dupe AI chatbots (16 Aug 2026)
  5. Wikipedia / secondary aggregation of FARA chain and volume metrics (updated Sep 2026)