NewsGuard AI Launches Vetted Chatbot to Combat AI Hallucinations
The new service replaces open-web scraping with a curated database of 12,000 reliable sources and a revenue-sharing model for publishers.
NewsGuard has launched NewsGuard AI, a specialized chatbot designed to deliver news responses drawn exclusively from journalist-vetted, reliable sources. The tool aims to eliminate the misinformation and "hallucinations" common in general-purpose AI by restricting its knowledge base to a curated whitelist.
Unlike traditional large language models that scrape the open web, NewsGuard AI draws its responses from approximately 12,000 news and information sources rated as reliable by NewsGuard journalists. To ensure accuracy, the company implemented 41 editorial safeguards, including a database of 64,000 debunked false claims used to prevent the repetition of misinformation. Additionally, the chatbot is programmed to refuse requests to write essays or reports for students.
The Trust Gap in AI
This launch follows a yearlong audit by NewsGuard which revealed a significant reliability problem among leading AI models. According to the audit, these models repeated false or misleading claims on controversial news topics 35% of the time—a sharp increase from 18% in August 2024 to 35% in August 2025.
NewsGuard has been rating the reliability of online news sources since 2018. The company's move into AI comes amid heightened tension between AI developers and the media industry, as publishers increasingly clash with tech firms over the unauthorized scraping of content without compensation.
A New Economic Model
Beyond technical safeguards, NewsGuard AI proposes a sustainable financial relationship with the press. The service employs a 50-50 revenue-sharing model with participating publishers whose journalism is cited, alongside affiliate-style subscription referrals.
This approach has already attracted partners who previously resisted AI integration. Chris Richmond, CEO of Snopes, noted that while his organization has restricted most AI chatbots from scraping its content, they are happy to partner with NewsGuard on a model that "does this the right way."
Industry Implications
By replacing the open-web training set with a human-vetted whitelist, NewsGuard AI attempts to solve the "trust gap" currently plaguing AI-generated news. The shift toward a direct payment model suggests a potential alternative to the adversarial relationship currently defining the AI-publisher landscape.
As the industry grapples with the proliferation of synthetic content, the ability to distinguish fact from fiction remains a primary concern. Nicholas Thompson, CEO of The Atlantic, emphasized the urgency of the issue, stating that few things will matter more in the near future than the ability of humans to determine what is real, what is false, and what is "confabulated nonsense."