One pipeline
Article intake, content extraction, signal review, and evidence packaging run as a single sequence — not independent steps stitched together.
Article text goes in. A structured bias analysis — direction, evidence, confidence, and next reads — comes out.
Paste article text or provide a URL. NeutralEye accepts both — URL submissions extract the article body automatically.
The system confirms the material is article-like. Short passages, navigation text, and blocked pages are flagged before analysis starts.
Tone, framing, attribution, source balance, and omission are reviewed together in one pass — not as five independent checks.
A confidence score is calibrated against how consistently signals appeared. Articles where all five checks point the same way score higher than those with mixed or thin evidence.
The result comes back with a direction label, confidence score, quoted evidence, and sources to read alongside — everything needed to check the reasoning yourself.
Article intake, content extraction, signal review, and evidence packaging run as a single sequence — not independent steps stitched together.
Every result uses the same schema: direction, confidence, summary, examples, sources, and recommendations — making results comparable across runs.
The system returns what it found and where. Quoted language and sourcing patterns are included so the output can be checked against the original text.
Every article goes through tone, framing, attribution, source balance, and omission in a single sequence — not as independent checks stitched together. The output stays tied to what was actually in the text.
Framing, source balance, attribution, and omission reviewed in one pass.
Summary, examples, confidence, and next-reading context stay attached.
Every result includes the direction label, a confidence score, quoted examples of the signals that shaped it, and sources to read alongside — so the analysis is a starting point, not a final word.
Readable explanation with evidence attached.
Every result includes a confidence score reflecting how consistently the detected signals appeared across the article. When evidence is sparse or ambiguous, the score drops — so you know when to read the analysis with more caution.