Israel has funded a multimillion-dollar campaign to dictate how AI chatbots, such as ChatGPT, Claude, Gemini, and Perplexity answer questions about Gaza, the Israel Defense Forces (IDF), and related topics and generating Israel propaganda, according to documents filed with the US Department of Justice (DoJ), reviewed by Politico, The Guardian, and TIME.
Three of six tested chatbots cited Israeli government-backed websites in answers about US-Israel relations, raising concerns that Israel propaganda could gain credibility without users knowing who funded those sources, according to a report published by WIRED Middle East.
The test showed that campaign websites reached some chatbot answers. It didn’t establish that the campaign changed the chatbots’ overall positions, entered their training data, or changed users’ views. The distinction matters as more people turn to chatbots for answers about contested political issues.
Which Chatbots Used the Campaign’s Websites?
WIRED Middle East put identical questions to ChatGPT, Claude, Perplexity, Microsoft Copilot, Gemini and Grok. All six described stronger US-Israel military cooperation as beneficial, although some added significant caveats. That shared answer alone cannot be attributed to the campaign. The test did not compare responses before and after the websites appeared online
The clearer finding concerned sources. Copilot, Claude and Grok cited Allyvia.org and repeated its claim that the US Israel alliance delivers $48 billion in measurable value. The site gave no supporting data for that calculation.
Claude told users to “take this figure with caution,” but the warning concerned the number, not the site’s funding. Copilot also cited Paxpoint.org in a separate answer about Israeli Palestinian peace efforts.
Perplexity did not use the $48 billion figure.
In another test, it recognized Allyvia in its search results and declined to cite it. WIRED Middle East did not report ChatGPT or Gemini citing the campaign sites in these tests. Nor did it report any of the six telling users, without being asked, that the cited websites were operated for an Israeli government-backed campaign.
That is the measured reach of Israel propaganda in the main investigation: campaign pages entered some answers as sources, while the effect on the answers’ broader conclusions remains unproven. A separate investigation described by Middle East Monitor found other chatbot citations under different questions, so its results should not be confused with the six-chatbot test.
US registration filings cited by WIRED Middle East show that Clock Tower X was contracted to receive $46.5 million in 2026 for Israeli-government-backed campaigns. The contract described an aim to influence how chatbot conversations frame issues. The company operated a network of sites presenting different pro-Israel arguments, including Allyvia’s economic claim and Paxpoint’s account of peace efforts.
Critics call such material Zionist propaganda because it advances a political case while its sponsorship can be easy to miss in a chatbot answer. The issue is attribution where a visible link tells readers where a claim appeared, but it may not tell them who financed the website.
Can Online Content Quietly Shape AI Answers?
The campaign shows how Israel AI propaganda can reach a chatbot through live search. A system may retrieve a page while preparing an answer and cite it as evidence. That can happen even if the page was never used to train the underlying model.
“Propaganda does not need to be in the training data,” told Carnegie Mellon University researcher, Zoey Jiang, to WIRED Middle East, adding that“it can be retrieved in search.”
Campaign pages appearing in the Common Crawl web archive therefore do not prove that any particular AI company trained on them.
AI and computational propaganda can also gain apparent credibility through repetition. Several polished websites may support the same claim while seeming independent of one another. A chatbot that brings those pages together could give users an impression of broad agreement without recognizing their common sponsor. That is a risk identified by researchers, not proof that every answer in this test was changed.
A second example came in August.
Daily Sabah, citing an investigation by Responsible Statecraft, reported that the Hanover Institute for Public Policy presented itself as a US style think tank while a disclaimer identified its creation on behalf of the Israeli Government Advertising Agency. Its reports used footnotes, statistics, and a neutral tone.
NewsGuard analyst Alice Lee called its appearance “a perfect mimicry of a typical credible American think tank.”
The “Hanover Institute for Public Policy” shows how Israel AI generated propaganda could appear credible to both readers and automated systems, although the reports provided here do not establish that its articles were AI generated or that they changed chatbot answers.
Claims about AI-generated propaganda need the same care. Content designed for AI discovery is not necessarily written by AI.
AI hasbara is Israel-focused public messaging directed at AI systems, but it’s the documented funding and sourcing that readers can check.
Describing every cited page simply as Zionist propaganda would leave out the more useful question. Did the chatbot explain the page’s origin before presenting its claims?
For AI companies, tracing ownership, checking funding notices and testing whether clusters of websites share a sponsor could help identify Israel using AI propaganda tactics and similar campaigns elsewhere. Readers can open a chatbot’s citations, inspect a site’s disclosures, and compare a disputed claim with independent reporting or primary records.
The evidence leaves Israel propaganda at a high threshold, some campaign claims appeared in chatbot answers, but the reports cannot show how much the campaign changed those answers or whether it persuaded users.
As AI and computational propaganda spread, source disclosure may determine whether a citation helps readers judge an answer or gives a sponsored claim of undeserved authority.
For now, AI-generated propaganda is an issue that requires specific evidence for each claim. The Israel AI propaganda isseen as a major alert for users as to how information is not being stated as it is, and this reminds us how history repeats itself with a twist of technology added to it.
The tested Israel propaganda websites demonstrate a concrete problem. A chatbot can name its source while leaving the source’s government connection out of the answer.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.
