Announcing AccuracyCheck: The First AI Hallucination Detector for Newsrooms

Journalists are stretched thin. AI can give them back time to focus on quality storytelling and community impact by taking care of routine tasks. It seems like a perfect match. 

However, without industry-specific safeguards, AI runs a high risk of delivering misinformation. To a news brand, even a minor inaccuracy in reporting is catastrophic.

Collaborator’s AccuracyCheck was built to solve that problem.

AccuracyCheck is the first AI hallucination detector built specifically for newsrooms. 

This new feature is now fully integrated into Magid’s Collaborator tool, adding an extra layer of quality assurance for journalists.

What’s an AI hallucination? It’s when AI generates content that sounds plausible but is factually incorrect. For journalists, this risk is dangerous.

Journalists need AI they can trust and rely on.

A BBC study analyzing ChatGPT, Copilot, Perplexity, and Gemini found that more than half of AI-generated answers to news-related questions had major issues, 19% introduced factual errors, and 13% featured altered or made up quotes. 

Contrastingly, Collaborator has a 0.0162% error rate. Now with AccuracyCheck, that rate is reduced to virtually zero.

AccuracyCheck goes far beyond basic AI evaluation. 

AccuracyCheck runs automatically whenever a user transforms an input into an output (e.g., turning a broadcast script into a web story). It functions as a safeguard, ensuring that nothing is created that didn’t originate from the journalist’s inputs. 

In rare cases when something does (only 0.0162% of Collaborator runs), AccuracyCheck flags it and provides a quick “Fix It” button for seamless corrections.

AccuracyCheck, Magid’s AI hallucination detector for newsrooms, provides journalists with:

• Near-zero misquote risk: It meticulously flags any potential misquotes, ensuring that original source material, like a direct quote, remains absolutely unchanged. For example, “Today was a beautiful day” will never morph into “Today’s weather was wonderful” without getting flagged to the user.
• Domain-specific fidelity: Unlike off-the-shelf solutions, AccuracyCheck is built on Magid’s seven decades of industry expertise. It understands the unique requirements of journalistic phrasing, brand compliance, and data synthesis.
• Easy correction: If an issue is flagged, AccuracyCheck provides a quick “Fix It!” button, allowing for seamless and immediate corrections, minimizing editorial overhead.
  

For news organizations, the ideal integration of AI means journalists can leverage its time-saving benefits while maintaining complete confidence in the veracity of content produced. Collaborator delivers this with a 99.98% success rate.

Ready to experience the Magid difference?

Every piece of content generated in Collaborator is automatically verified to ensure it adheres to the journalist’s ground truth. We invite you to build your AI confidence with Magid-approved AI-assisted publishing, where time-savings meets uncompromising journalistic integrity.

Interested to learn more about Collaborator? Let’s talk.

Existing user? Log in.