WASHINGTON (Azat TV) – Major artificial intelligence (AI) platforms, including ChatGPT, Claude, and Gemini, are frequently citing foreign propaganda sources when responding to user queries about international conflicts, with a significant prevalence in coverage of the Israel-Hamas conflict. This concerning trend, detailed in a new study released on March 3, 2026, by Foundation for Defense of Democracies’s Center for Cyber and Technology Innovation (CCTI), highlights a growing national security threat and undermines the integrity of information accessed by millions.
The CCTI research, conducted between October and November 2025, involved querying approximately 180 questions across three major conflicts: Israel-Hamas, Ukraine-Russia, and Taiwan-China. The study found that state-aligned propaganda appeared in a staggering 57 percent of all AI responses. These platforms often direct users to sources like Al Jazeera, Anadolu Agency, Pravda, and China Daily, even when questions are framed to elicit answers unfavorable to U.S. adversaries.
AI Platforms and Propaganda in Gaza Coverage
The study revealed particularly stark findings regarding the Israel-Gaza conflict. Al Jazeera, a Qatar-based network identified by CCTI as state-aligned due to its funding structure and lack of editorial independence from the Qatari government, emerged as the third-most cited source overall, behind only Wikipedia and Congress.gov. LLMs directed users to Al Jazeera in 52 percent of all questions asked, totaling 228 citations.
For questions specifically about the Gaza conflict, Al Jazeera citations appeared in 70 percent of neutral queries. More remarkably, the network was cited in 65 percent of questions explicitly biased in favor of Israel. This suggests Al Jazeera’s success in Generative Engine Optimization (GEO), optimizing content to increase its visibility within AI-generated responses, rather than indicating pro-Israel coverage, according to CCTI.
Beyond Al Jazeera, the study also identified other problematic sources frequently cited by AI models in relation to the Israel-Gaza conflict. These included Turkey’s state-run Anadolu Agency, which appeared in 37 percent of all Israel-Gaza conflict questions. Other Turkish media sites like Yeni Safak, Daily Sabah, and A News, which lack editorial independence from the government in Ankara, were also cited. ChatGPT, in particular, cited The Palestine Chronicle, an outlet currently under congressional investigation for potential ties to Hamas and Iranian state media, and Ninkilim, a pro-Palestinian activist website known for posting calls for violence.
A particularly alarming discovery was ChatGPT’s citation of Al Mayadeen, identified as a mouthpiece for Hezbollah, across all three conflicts studied, including queries about the Israel-Gaza conflict. Traditional Google searches for the same questions yielded no Al Mayadeen hits within the top 100 results, underscoring the unique vulnerability of LLMs to such sources.
The Mechanics Behind AI’s Propaganda Problem
CCTI attributes the prevalence of state-aligned media in AI citations to several factors. AI training models rely heavily on outlets characterized by high publication volume, global reach, and accessibility. These attributes are precisely what influential state-aligned propaganda outlets possess. In contrast, premium news sources in democratic nations often operate behind paywalls or block AI crawlers, preventing their content from being used for AI model training or cited in responses.
Many major legacy media sites actively block LLM crawlers to prevent unauthorized use of their content, with some, like The New York Times, even suing AI developers for content use. This creates a structural advantage for state media outlets from countries like Qatar, Russia, Turkey, and China, whose content is freely accessible and thus disproportionately represented in the vast web-crawl datasets used by AI platforms.
The study also highlighted a critical concern: when LLMs cited state-aligned media in response to neutral questions, the linked articles often contained content that was significantly more extreme or biased than the AI’s initial answer. This creates a ‘pipeline to more extreme views’ for users who click through these citations for verification, undermining the very purpose of source attribution and posing a serious national security vulnerability.
Implications for National Security and Independent Media
The deep reliance of AI tools on problematic sources for citations presents significant national security risks. As AI increasingly serves as a primary information source for students, educators, and government agencies, the absence of robust guardrails against adversarial foreign content can subtly shift narratives and expose users to biased or extremist viewpoints. The White House, through its AI Presidential Challenge, is actively promoting the deployment of AI tools across federal agencies and K-12 schools, making citation integrity a crucial component of AI safety frameworks.
Furthermore, the economic ramifications for independent news organizations are dire. In the digital media landscape, traffic driven by citations translates into subscriptions and advertising revenue. By preferentially directing traffic to state-aligned propaganda outlets, AI tools are creating structural advantages for these entities while eroding the financial viability of independent journalism. If this trend continues, AI could play a pivotal role in strengthening state-aligned propaganda networks globally.
To address these flaws, CCTI recommends that AI vendors actively deprioritize state-controlled media in their outputs. Federal procurement standards should demand transparency in citation patterns, and citation integrity must be integrated as a core component of AI safety frameworks. This proactive approach is deemed essential to prevent the technological future from being built on a compromised informational foundation.
The findings underscore a critical challenge in the evolving digital information landscape, where the widespread adoption of AI tools, without adequate safeguards, risks inadvertently legitimizing and amplifying narratives aligned with foreign state interests, particularly concerning sensitive geopolitical events like the Israel-Gaza conflict.