The fake US thinktank first exposed last week has now been quantified by the Guardian: 124 reports and 560,000 words in nine days, built on a commercial platform that promises to optimize content so AI chatbots will cite it — with an Israeli government funding trail documented in US Justice Department filings. It is the clearest public example yet of state-linked actors targeting the AI answer layer itself: not readers, but the models that increasingly speak for them. It is the kind of story our AI industry coverage tracks as chatbot manipulation goes industrial. (We covered the site's initial discovery here.)
A thinktank that does not exist
The site publishes under the name of the Hanover Institute for Public Policy. According to the Guardian's Jason Wilson, the institute apparently does not exist as a legal entity in any jurisdiction, has no physical address, and carries no named staff and no bylines on any of its reports. Its own terms of use are governed by the laws of "the state in which the Institute is established" — a state the document declines to name.
The scale of output was industrial. Between August 6 and August 14, the Hanover Institute published 124 reports totaling more than 560,000 words, an average of about 4,500 words each. On August 12 and 13 alone it produced 73 reports amounting to almost 354,000 words. Nothing has been published since the first media reports about the site appeared.
Engineered for the chatbot answer layer
What marks the operation as novel is its targeting. All but one of the 124 headlines opens with a question resembling one a user might type into a chatbot — "Is Anti-Zionism Antisemitism?", "Did Israel Expel Palestinians From Their Land?", "Is Israel Carrying Out a Genocide in Gaza?" The reports present Israel's position on torture of Palestinian prisoners, war crimes, and starvation in Gaza as neutral research, formatted for machine retrieval.
The mechanics are documented in official filings. The New York media production company Piro Inc registered the site with the US Justice Department this month under the Foreign Agents Registration Act (Fara), a 1938 law requiring disclosure of work for foreign governments, as material it distributes for the Israeli government. The Guardian reports the effort is financed by tens of millions of dollars from the Israeli government, routed through third parties including the European advertising group Havas Media. A contract between Havas and another American firm working on the campaign commits it to the "deployment of websites and content to deliver GPT framing results on GPT conversations."
After the site's existence was first reported on August 14, the institute quietly added a funding page containing a telling correction: "An earlier version of this page said the Institute operated without external funding, and that was true when it was written." The Guardian contacted Piro Inc, Havas Media, and the Israeli government advertising agency LaPam for comment.
Why chatbots are the new battlefield
The strategy has a name in the marketing industry — generative engine optimization, or GEO — and it is the successor discipline to search engine optimization. When a chatbot answers a question, it draws on its training corpus and retrieval sources. Content that is voluminous, question-formatted, and styled as authoritative research is more likely to be surfaced, cited, or absorbed into a model's answer. A fake thinktank is nearly the ideal GEO artifact: institutional credibility without an institution.
The timing reflects desperation as much as sophistication. The Guardian notes the campaign comes as Israel's standing in US public opinion has collapsed, and as Israeli officials brief media outlets that money spent on "Hasbara" public diplomacy overseas has been wasted.
The Hanover Institute is not an isolated case. The same day the Guardian published its investigation, OpenAI announced it had disrupted a new covert Russian influence campaign using AI — the latest in a series of takedowns the company has attributed to state-linked actors. Two separate operations, two continents, one shared insight: influencing models is now cheaper and more durable than influencing people directly.
The governance gap
For regulators, the case exposes a hole in existing frameworks. Fara disclosure caught the operation on paper, but only after the content was live and potentially ingested. The EU AI Act's transparency provisions and the voluntary frameworks governing training data provenance were not designed for a world where the cheapest path to a population's beliefs runs through its chatbots.
For AI labs, the operational question is immediate: how do retrieval systems and training pipelines distinguish research from content farms wearing research clothing? Volume, formatting, and confident citations — the signals of authority — are exactly what the Hanover Institute mass-produced. Until models get better at provenance, the half-million words are already out there, and no disclosure filed after the fact recalls them.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
