Israel Contracts Firm for AI Content Campaign

Israel's government hired US firm Clock Tower X in September 2025 under a reported $6 million contract to produce online content and websites intended to influence how GPT-style chatbots frame Israel-related topics. Responsible Statecraft reported that the work included Gen Z-focused distribution and search optimization using MarketBrew AI; Anadolu Agency later described a separate reported $46.5 million campaign involving pro-Israel sites.
Israel's government hired US firm Clock Tower X under a reported $6 million contract to create digital content and websites intended to influence how GPT-style AI systems frame Israel-related subjects, according to Responsible Statecraft. The outlet reported on September 29, 2025 that the contract calls for Clock Tower to deploy "websites and content to deliver GPT framing results on GPT conversations."
The reported work is notable because it targets the information environment surrounding generative AI rather than only conventional social-media distribution. Responsible Statecraft reported that at least 80% of Clock Tower's content was to be tailored to Gen Z audiences across TikTok, Instagram, YouTube, podcasts, and other digital and broadcast outlets, with a minimum target of 50 million impressions monthly. Its account identifies former Trump campaign manager Brad Parscale as the leader of Clock Tower X.
What the contract describes
According to Responsible Statecraft, Foreign Agents Registration Act filings describe Clock Tower's assignment as helping "execute a nationwide campaign in the United States to combat antisemitism." The filing did not specify the messages to be promoted. The New Arab likewise reported that public contracts did not detail the content to be promoted on Israel's behalf.
Both outlets reported that the work includes use of MarketBrew AI, a search-engine optimization platform. Responsible Statecraft characterized the intended use as improving the visibility and ranking of relevant narratives in search engines, while The New Arab reported that the tool could help clients adapt content for algorithmic ranking.
An Anadolu Agency report published July 29, 2026, citing Drop Site News and contract documents, described a larger reported $46.5 million Israeli government campaign involving Clock Tower X. Anadolu reported that the effort produced hundreds of articles across multiple sites and named Allyvia.org, FactSignal.org, and Paxpoint.org among sites promoting pro-Israel perspectives. The report did not establish that the $46.5 million arrangement and the previously reported $6 million contract are the same agreement.
Training data versus chatbot retrieval
The available reporting describes an attempt to influence chatbot outputs through web publishing, search indexing, and potential inclusion in web datasets. It does not demonstrate that OpenAI, Anthropic, Google, Microsoft, or Perplexity changed a model's training data, ranking systems, or product behavior in response to the campaign.
Anadolu, again citing Drop Site, reported that some linked sites had been referenced in responses from Perplexity and Microsoft Copilot, and that the sites had entered Common Crawl, a public web archive commonly used in language-model data pipelines. Common Crawl inclusion alone does not establish that any particular model was trained on the pages, nor does it establish their weight in a training corpus.
For ML practitioners, the distinction matters. Retrieval-augmented and web-search-enabled systems can surface recently indexed material more quickly than a model's underlying parameters change. Across the industry, attempts to shape web-scale corpora and search results reinforce the value of provenance signals, source-quality controls, adversarial evaluation, and citation inspection for politically sensitive queries.
Haaretz disinformation and cyber correspondent Omer Benjakob, speaking on a Haaretz podcast published in April 2026, described the sites as designed to resemble research institutes and think tanks and to optimize visibility for ChatGPT, Gemini, Claude, and other chatbots. He questioned the campaign's efficacy and value. The reporting reviewed here does not provide independent measurements of its impact on answers from those systems.
Key Points
- 1Reported Clock Tower X contracts explicitly include websites intended to shape GPT conversation framing, extending state messaging into generative-AI information channels.
- 2Common Crawl presence and chatbot citations can affect retrieval or future datasets, but neither proves direct model-training influence or durable response changes.
- 3For AI teams, politically contested web content underscores the need for provenance checks, source diversity, and adversarial testing in search-grounded systems.
Scoring Rationale
The reported contracts raise material information-integrity questions for teams building or evaluating web-grounded LLM applications. Evidence of direct changes to frontier-model training or behavior is not established, which limits the score relative to a confirmed platform-level manipulation incident.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems