How NarrativePrism works
Our analytical methodology — from raw article to structured narrative intelligence.
From articles to atoms
Every article that enters NarrativePrism is decomposed into its structural components:
Claims
Specific, atomic assertions extracted from the text. Each claim is a discrete statement that can be verified, contested, or tracked independently. Claims are linked to their supporting evidence — direct quotes, data citations, attributed statements.
Frames
The analytical lens an article applies to its subject. Frames reveal what's being emphasized and what's being deemphasized. The same event can be framed as a safety concern, an economic opportunity, or a political strategy — and each framing shapes the reader's understanding differently.
Angles
The article-level storytelling strategy. Where frames are reusable lenses that apply across stories, angles are specific to a single piece of coverage — the particular combination of emphasis, sourcing, and narrative structure a journalist chose.
Narratives
Higher-order patterns that emerge when claims, frames, and voices align. Narratives are tracked as competing entities — we never blend opposing narratives into a single summary. When three outlets say “regulation is necessary” and two say “regulation will stifle innovation,” you see both narratives separately, with their supporting evidence.
Voice attribution
Traditional media monitoring attributes information to “sources” or “reports.” NarrativePrism attributes every claim to a specific voice.
What a voice profile includes:
- Identity
- Name, role, affiliated outlet(s)
- Topic focus
- What they typically cover and their areas of expertise
- Publishing cadence
- How frequently they publish and in what patterns
- Editorial lean
- Their positioning on the political spectrum, based on their body of work
- Citation patterns
- Who they cite, who cites them, and how these relationships evolve
- Drift detection
- Automated flagging when a voice's position or framing departs significantly from their established patterns — a signal that something may be shifting
Frame analysis
NarrativePrism tracks how the same event or topic is framed differently across outlets and over time.
What frame analysis reveals:
- Emphasis mapping
- What aspects of a story each outlet highlights — economic impact, human interest, political implications, technical details
- Omission detection
- What one outlet covers that another ignores. Omissions are often more informative than inclusions
- Spectrum positioning
- How framing distributes across the political spectrum (left, center, right) using established classification methodologies
- Coherence analysis
- How internally consistent a narrative is across its supporting claims and sources. Incoherent narratives — where the evidence doesn't fully support the frame — are flagged
Multi-perspective coverage
NarrativePrism monitors sources across the full ideological and institutional spectrum:
Political spectrum
Left, center, right, and non-aligned outlets
Institutional type
Legacy media, digital-native, independent, wire services, trade publications
Geographic scope
National, regional, international
This breadth enables:
- Cross-source comparison
- How Fox News, CNN, The New York Times, and The Intercept cover the same event — structurally, not just tonally
- Omission mapping
- What Reuters reported that The Washington Post didn't mention
- Amplification tracking
- Which outlet's framing gets picked up and repeated by others
- Narrative velocity
- How quickly a frame spreads across the source landscape
Domain specialization
Media coverage of AI policy requires different sources, different voices, and different context than coverage of NFL trades. NarrativePrism provides domain-specific analytical environments:
- Focused environments
- Politics, Tech, Business, Sports, and more — each with curated source lists, pre-tracked voices, and domain-relevant topic structures
- Source curation
- Each environment monitors the outlets that matter for that domain — tech blogs for Tech, financial publications for Business, political magazines for Politics
- Voice scoping
- A tech reporter appears in your Tech environment. A political correspondent appears in Politics. No cross-domain noise
- Domain context
- Analysis is informed by domain-specific knowledge — what counts as a significant claim in biotech is different from what counts in electoral politics
Continuous analysis
NarrativePrism operates as a continuous analytical pipeline, not a batch processing system:
- Ingestion
- Articles are sourced from hundreds of outlets, parsed, deduplicated, and prepared for analysis within minutes of publication
- Extraction
- AI models extract claims, identify voices, detect frames, and map narrative relationships on every article
- Evolution tracking
- Narratives are tracked over time. When a narrative accelerates, declines, or shifts framing, the change is detected and surfaced
- Shift detection
- We monitor for shifts at every level — from individual quote emphasis to article-level angles to topic-wide narrative reframing
- Progressive disclosure
- Analytical depth is pre-computed. When you click into a narrative, the supporting claims, voices, and evidence are already there — no re-prompting, no waiting for a new AI query