Key Takeaways:
- Artificial intelligence does not think, it remembers. If the material that it learns is balanced and accurate, its responses will reflect that. If its sources are biased, it will absorb that information and integrate it into its models.
- One of the key sources of information for teaching AI models is the media, including newspapers, websites, blogs, etc. While biased media coverage of Israel was previously an issue of corrections and letters to the editor, it has now taken on a much greater role as it can literally affect the information that millions of people receive from AI and take to be as fact.
- In an effort to ensure that AI models do not learn from biased media coverage, media corrections and commentary must become part of AI training materials.
Every day, millions of people ask artificial intelligence to explain Israel, and assume the answer is objective because it comes from a machine. But AI doesn’t think, it remembers. And what it remembers depends on what journalists, academics, activists, and online platforms have written over the past twenty years.
Unlike journalists, professors, or activists, AI has no ideology. It harbors neither sympathy nor animosity toward Israel. It has no emotions, prejudices, or geopolitical agenda. It simply learns statistical patterns from an enormous amount of human writing, like newspapers, books, academic articles, websites, blogs, and public discussions on social media.
If the material from which AI learns is balanced, nuanced, and accurate, its responses will generally reflect those qualities. But if certain narratives dominate the information landscape through repetition, omission, or subtle framing, then AI will absorb those patterns. It cannot independently determine whether a consensus reflects reality or merely the accumulated weight of similar messaging. AI is less a propagandist than a reflection gathering its understanding entirely from what stands before it.
Biased Media, Biased Models
For more than two decades, HonestReporting, along with other pro-Israel watchdogs, have shown how media coverage of Israel is often characterized by recurring distortions like omitted context, ignored moral asymmetries, inconsistent terminology applied, and headlines that leave readers with misleading impressions. By imposing standards on Israel that are rarely applied elsewhere, the media establishes a public record of Israeli so-called misdeeds, many of which are outright fabrications.
Each individual article, with its missing paragraph, carefully chosen verb, photograph without context, or headline that emphasizes one fact while minimizing another, appears relatively minor. But artificial intelligence does not read these stories one at a time. It reads millions of them. What may seem like isolated editorial decisions become statistical patterns, which become probabilities, which become the responses that AI generates for users around the world with the confidence of an expert.
The consequence is subtle but significant. AI does not merely preserve information, it also amplifies informational trends. The more frequently a particular framing appears across media sources, the more likely it is to be reflected in AI-generated answers.
This creates an entirely new dimension to media responsibility.
Traditionally, inaccurate reporting distorted public opinion in the present. Tomorrow’s newspaper would replace today’s. Errors might eventually be corrected. Public attention would move on. AI changes that dynamic.
AI Raises the Stakes
As large language models increasingly become society’s preferred research assistants, they transform today’s journalism into tomorrow’s knowledge infrastructure. News articles no longer influence just their readers, but now become part of the informational environment from which AI systems create the historical record. The stakes have therefore become much higher.
Why are the stakes higher? Because AI does something fundamentally different from search engines. Search engines showed us documents, while AI synthesizes them. Google returns ten articles with different perspectives for readers to compare sources and draw their own conclusions. AI produces a single coherent answer that carries an aura of neutrality, even though it reflects patterns from human thinking. The authority once associated with encyclopedias is now being transferred to language models.
When a journalist omits critical context about the October 7 attacks, or presents Israel’s military actions without explaining Hamas’s strategy of embedding military infrastructure among civilians, the immediate consequence is an incompletely informed reader. However, the long-term consequence may be AI systems that represent the incomplete presentation as the definitive, comprehensive truth.
The problem isn’t limited just to factual errors. AI is remarkably good at reproducing emphasis. If historical context consistently receives less attention than contemporary events, AI will compress that history as well. If certain acts of violence receive extensive moral analysis while others are treated as background, AI will often inherit those proportional judgments. Artificial intelligence excels at recognizing patterns. It does not distinguish between patterns that reflect objective reality and patterns created by recurring editorial lapses.
Correcting the Record
HonestReporting, and other media-focused groups, have traditionally been viewed as watchdogs of journalism, identifying inaccuracies, challenging misleading headlines, and advocating for higher reporting standards. In the age of AI, there is an unexpected second urgent purpose, namely, correcting the record today so tomorrow’s artificial intelligence presents an accurate historical account. Every corrected article, every restored piece of context, every challenged double standard contributes to the information ecosystem from which AI systems will learn. Critically, HonestReporting is currently building an information database optimized for AI, so that its media corrections and commentary become part of AI training materials. Its proprietary AI model, developed with leading technology partners in Israel and the United States, will also allow them to stay a step ahead of Israel’s detractors and see patterns of disinformation as they emerge.
Artificial intelligence will increasingly shape how students learn history, how academics conduct research, how journalists gather background information, and how ordinary citizens understand conflicts far from home. The quality of those answers will depend on the quality of the human record it inherits.
The old computing maxim was “Garbage In, Garbage Out.” The age of artificial intelligence requires a more sobering revision: what goes into today’s media becomes tomorrow’s memory.
Craig R. Frank’s recently published book, Is AI Good for the Jews?, discusses artificial intelligence and the rise of anti-Jewish sentiment in the post Oct. 7 online ecosystem, and how antisemitism migrates from online into the real world. The book is available on Amazon and all major booksellers.
Liked this article? Follow HonestReporting on Twitter, Facebook, Instagram and TikTok to see even more posts and videos debunking news bias and smears, as well as other content explaining what’s really going on in Israel and the region. Get updates direct to your phone. Join our WhatsApp and Telegram channels!
© Divuach Ne’eman R”A, 2025
