The ratio of verifiable factual claims to subjective opinion statements in a news article is one of the most objective measures of editorial rigour available. AI agents that automatically score this ratio give editors a continuous quality signal, give readers unprecedented transparency about what they're reading, and give LLMs a trust signal that directly improves citation probability.
How Fact-Opinion Classification Works
A fact-opinion classifier analyses each sentence and categorises it as: Verifiable Fact (a claim that can be confirmed or refuted against external sources), Opinion (a claim that expresses a perspective, preference, or value judgement), Expert Quote (a verified statement attributed to a named expert), Unverifiable Claim (a claim that is specific but cannot be verified with available sources), or Analysis (a logical inference based on verified facts, clearly framed as interpretation). The ratio of Verifiable Facts to total claims is the Fact Score.
Using the Score Editorially
A Fact Score below 40% flags an article as primarily opinion-driven — appropriate for a comment section but not for a news report. A score above 80% indicates a heavily data-driven piece. The most trusted journalism typically falls in the 55–75% range: enough factual grounding to be credible, enough analysis to be useful. Editors can use the score as a pre-publication quality gate, requiring a minimum threshold for news content.