Political Bias Rating

Sources are rated on a -3 to +3 scale based on assessments from Media Bias/Fact Check, AllSides, and Ad Fontes Media.

-3 Far Left  ·  -2 Left  ·  -1 Lean Left  ·  0 Center  ·  +1 Lean Right  ·  +2 Right  ·  +3 Far Right

Reliability Rating

Sources are rated 1–5 for factual reliability based on track record, corrections policy, and sourcing standards.

5 Very High (wire services)  ·  4 High  ·  3 Mixed  ·  2 Low  ·  1 Very Low

Methodology

Intelligence Community Standards for News Analysis

BLUFify Analytical Standards

BLUFify™ applies Intelligence Community analytical standards to news aggregation. Our methodology adapts principles from:

  • ICD 203: Analytic Standards for objectivity, independence, and timeliness
  • ICD 206: Sourcing Requirements for transparency and traceability
  • ICD 208: Writing for Maximum Utility (customer focus)
  • IFCN: International Fact-Checking Network principles
  • SPJ: Society of Professional Journalists Code of Ethics

This unified framework ensures every briefing is accurate, transparent, and actionable.

Eleven Unified Analytical Standards

1

Source Credibility

Every source rated for bias (-3 to +3) and reliability (1-5). Ratings based on Media Bias/Fact Check, AllSides, and Ad Fontes Media.

2

Confidence Levels

Every assessment rated HIGH, MODERATE, or LOW confidence based on source quantity, reliability, and corroboration.

3

Source Transparency

Every briefing includes a Source Summary Statement with source count, reliability breakdown, and political spectrum coverage.

4

Fact/Analysis Separation

Clear distinction between verified facts and analytical judgments. Facts are attributed; analysis is labeled.

5

Cross-Spectrum Validation

Stories covered across political spectrum receive validation bonus. Single-spectrum stories flagged as potential blindspots.

6

Corrections Policy

Errors corrected promptly with transparent disclosure. All corrections logged publicly with original text preserved.

7

AI Transparency

AI-assisted analysis clearly disclosed. Claude AI summarizes and analyzes; humans validate and edit.

8

Customer Focus

Briefings prioritize actionable intelligence. BLUF format ensures key takeaways are immediate and clear.

9

ACH Alternative Hypothesis

Every strategic assessment includes a structured competing hypothesis per ICD 203 (Analysis of Competing Hypotheses). Named hypothesis, supporting evidence, and confirmatory indicator — not vague speculation.

10

Longitudinal Context Memory

Each briefing is informed by a rolling 5-briefing context window. Watch items, key questions, and tracked developments carry forward, preventing recurring events from being treated as new speculation each cycle.

11

Adaptive Topic Diversity

Automated Union-Find clustering detects dominant topics and enforces a per-topic cap on key-development slots. No single event — however significant — can crowd out coverage of other important issues.

Ethical Intelligence Network (EIN)

The EIN is our transparent framework for sourcing, accessing, and attributing news content. Every source is assigned a partnership tier that determines access method, rate limiting, and fallback priority.

Live Source Metrics

89
Total Sources
4
T1 Partners
10
T2 Syndicate
75
T3 Public
4.1
Avg Reliability

T1 Partner

4

Direct API access. Wire services with formal partnerships (AP, Reuters, AFP).

Rate limit: 500ms

T2 Syndicate

10

Licensed syndication. Major outlets with RSS agreements (BBC, NPR, PBS).

Rate limit: 1000ms

T3 Public

75

Public RSS feeds with attribution. Includes fallback chain support.

Rate limit: 2000ms

Ethical Fallback Chain

When a primary RSS feed is unavailable, our system follows a responsible fallback chain to maintain coverage without overloading source servers:

RSS Feed (Primary)
Archive.org
Skip (No Overload)

Rate limiting is enforced per tier to respect each source's infrastructure. Tier 1 partners are never accessed via fallback.

Source Coverage Balance

Our source directory is balanced across the political spectrum:

Left 25%
Center 53%
Right 22%

22 left-leaning · 47 center · 20 right-leaning sources

Browse all 89 sources →

Confidence Levels (ICD 203)

Following Intelligence Community Directive 203 (Analytic Standards), every assessment includes an explicit confidence level:

HIGH CONFIDENCE

Criteria: 3+ sources with cross-spectrum agreement, high reliability average (4+), primary reporting present, multiple corroboration points.

MODERATE CONFIDENCE

Criteria: 1-2 credible sources, plausible interpretation, limited corroboration, or single high-reliability source with original reporting.

LOW CONFIDENCE

Criteria: Single source, breaking/unverified, known information gaps, conflicting reports, or untested assumptions.

ICD 203 Probability Language

Confidence levels correspond to defined probability ranges and standard analytic language:

Level Probability Standard Language
HIGH ≥70% almost certainly, highly likely, we assess
MODERATE 40–69% possibly, may, could, we believe
LOW <40% unlikely, cannot confirm, we cannot rule out

ACH Alternative Hypothesis (ICD 203)

Every BLUFify™ strategic assessment includes a structured Analysis of Competing Hypotheses (ACH) — a core IC tradecraft standard from ICD 203. This is not generic "devil's advocate" framing; it is a named, evidence-grounded competing explanation for the primary analytical judgment.

ACH Format

Every alternative hypothesis follows a three-part structure:

  1. Named Hypothesis: A specific, falsifiable alternative explanation (not "consider the other side")
  2. Evidence: What observable facts better support this hypothesis than the primary judgment
  3. Indicator: A concrete, observable event that would confirm or refute this hypothesis

Example

"Diplomatic Signal Hypothesis: The military buildup may be coercive signaling rather than preparation for offensive action. Evidence: Back-channel diplomatic contacts reported by two regional outlets suggest negotiation tracks remain open. Indicator: Watch for a ceasefire proposal at the upcoming UN Security Council session — if tabled, this hypothesis strengthens significantly."

The ACH framework prevents analytical groupthink by requiring explicit engagement with the strongest competing interpretation — not just the most convenient one.

Longitudinal Context Memory

BLUFify™ maintains a rolling 5-briefing context window that feeds directly into each new strategic assessment. This prevents a critical failure mode common in real-time news systems: treating a known, ongoing situation as new speculation every cycle.

What Carries Forward

  • Watch items from prior briefings
  • Standing key questions
  • Top tracked development titles
  • Recurring event patterns

What This Enables

  • Contextually grounded assessments
  • Calibrated confidence over time
  • Detection of changing dynamics
  • Reduced redundant speculation

Practical Example

An ongoing conflict that has appeared in 5 consecutive briefings is analyzed with the accumulated context of those prior assessments — not treated as a novel development requiring baseline speculation.

Adaptive Topic Diversity Enforcement

A single high-profile event can generate dozens of stories across our 176+ sources, creating the risk of briefing saturation — where one topic dominates at the expense of other significant developments. BLUFify™ prevents this automatically.

How It Works

Our adaptive clustering algorithm uses Union-Find with Jaccard similarity on keyword fingerprints to detect topic families in real time — no manually maintained keyword lists required.

  1. Fingerprint: Each story cluster is reduced to a set of significant keywords
  2. Cluster: Stories with ≥25% keyword overlap are grouped into the same topic family
  3. Cap: Each topic family is limited to a maximum of 2 key-development slots in the briefing
  4. Fill: Vacated slots are filled with the next highest-priority story from a different topic

This process is self-calibrating. When a new major event emerges, it is detected and capped automatically. When an event subsides and fewer stories appear, it naturally recedes in the rankings without requiring any configuration change.

ICD 190 CRITIC Breaking News System

BLUFify's breaking news pipeline applies the Intelligence Community's ICD 190 CRITIC (Critical Intelligence Communications) standard to determine when an event warrants immediate notification. Breaking alerts are not issued for trending headlines or developing stories — only for events that would independently qualify for a CRITIC report.

Seven CRITIC Categories (ICD 190)

Category Examples
A — Hostile Act Military strikes, invasion, armed attack, declaration of war
B — Terrorist Act Mass-casualty attacks, active shooters, coordinated bombings
C — Political Disruption Coups, government collapse, martial law, constitutional crisis
D — Humanitarian Crisis Pandemic declarations, mass displacement, catastrophic disasters
E — Environmental Crisis Magnitude 7+ earthquakes, tsunami warnings, Category 4–5 hurricane landfall
F — Economic Crisis Market circuit breakers, sovereign defaults, emergency Fed action
G — Cascading Event Strait closures, NATO Article 5 invocation, national emergency declarations

Two-Stage Evaluation Pipeline

  1. Regex Pre-Screen: Every story from Tier 1 sources (AP, Reuters, Bloomberg, AFP, BBC, USGS, NHC, CDC, FEMA, WHO, SEC) is evaluated against 60+ CRITIC keyword patterns. Non-matching stories are discarded in <1ms. Overly broad exclusions are intentionally avoided — a false negative (missed CRITIC event) is worse than a false positive passed to Claude.
  2. Claude Sonnet Evaluation: Stories passing regex pre-screen are sent to Claude Sonnet 4.6 with ICD 203 confidence calibration instructions and longitudinal context (recent alert titles). Claude makes the final CRITIC determination, assigns a confidence level (HIGH/MODERATE/LOW), and provides a one-sentence reasoning statement.

Alert Fatigue Controls

  • Event deduplication: Jaccard similarity clustering groups all wire reports on the same event into one cluster — one alert per event, not one per source
  • Longitudinal suppression: Claude sees recent alert titles and distinguishes genuine escalation from routine continuation of known events
  • Confidence-aware gap: HIGH confidence alerts fire immediately; LOW/MODERATE alerts observe a 2-hour minimum inter-alert gap
  • 24-hour cap: Maximum 6 alerts per 24-hour window
  • Per-cycle limit: At most 2 distinct events alerted per 15-minute cycle (prevents single-cycle flood during concurrent crises)

Source thresholds: HIGH/MODERATE confidence requires one Tier 1 wire source. LOW confidence requires two wire sources. Authority sources (USGS, NHC, CDC, etc.) always qualify alone regardless of confidence level.

Seven-Factor Priority Scoring

Every story is scored using seven weighted factors (normalized to 0-100). Weights are transparent and tunable:

Impact (25%)

Potential consequences and scope. Higher impact means more people, institutions, or systems affected.

Urgency (20%)

Time-sensitivity with a 6-hour half-life. Breaking events and developing stories score highest.

Prominence (15%)

Coverage breadth and editorial placement. Cross-spectrum validation boosts prominence.

Source Quality (15%)

Reliability and factuality ratings of contributing sources. Wire services and high-reliability outlets score highest.

Novelty (10%)

Freshness compared to 7-day lookback. Genuinely new developments score higher than ongoing coverage.

Coverage Breadth (10%)

Number of independent sources covering the story. 3+ sources triggers a coverage bonus.

Relevance (5%)

User sector preferences and personalization. Stories matching subscribed sectors receive a boost.

Weights are exposed via our methodology API endpoint.

BLUFify Priority Framework (BPF)

Priority scores map to four formal tiers used throughout the platform:

Tier Score Range Description
BPF-1 CRITICAL ≥85 ICD 190 CRITIC threshold — immediate breaking alert
BPF-2 HIGH 70–84 Lead story priority in scheduled briefings
BPF-3 MEDIUM 50–69 Standard briefing inclusion
BPF-4 STANDARD <50 Background monitoring; may appear in digest

Source Credibility Rating

Every source is evaluated on three dimensions:

Political Bias Scale (-3 to +3)

-3 Far Left -2 Left -1 Lean Left 0 Center +1 Lean Right +2 Right +3 Far Right

Reliability Rating (1-5)

●●●●● Very High (5): Wire services, exceptional accuracy, rigorous sourcing
●●●●○ High (4): Strong factual reporting, corrections policy, minor issues rare
●●●○○ Mixed (3): Generally accurate but occasional concerns or opinion mixing
●●○○○ Low (2): Frequent errors, sensationalism, or misleading content
●○○○○ Very Low (1): Unreliable, significant factual issues, or propaganda

Factuality Score (1-5)

Based on IFCN (International Fact-Checking Network) principles. Measures track record of accurate, verifiable reporting independent of political perspective.

Ratings derived from Media Bias/Fact Check, AllSides, Ad Fontes Media, and internal analysis.

Source Summary Statement (ICD 206)

Every briefing includes a Source Summary Statement at the top. This provides immediate transparency about the sources informing the analysis.

Example: "Based on 12 sources (8 high-reliability) across political spectrum (4 left, 3 center, 5 right). HIGH CONFIDENCE: Corroborated by multiple independent sources with primary reporting."

BLUF Format

Bottom Line Up Front is a military and intelligence community standard for presenting information. The key conclusion comes first, followed by supporting details.

Every BLUFify™ briefing follows this structure:

  1. BLUF: The single most important takeaway
  2. Source Summary: Transparency on sourcing (ICD 206)
  3. Key Judgments: 3-5 analytical assessments with confidence levels
  4. Facts vs. Analysis: Clear separation per ICD 203
  5. Sources: Full list with credibility ratings

AI Transparency Disclosure

BLUFify™ uses AI assistance in the following ways:

  • Summarization: Claude AI (Anthropic) generates initial summaries from source articles
  • Analysis: AI assists in identifying patterns, connections, and key themes
  • Scoring: Seven-factor priority scoring is algorithmic with human oversight
  • Validation: All AI-generated content is subject to editorial review

AI assistance is a tool, not a replacement for human judgment. Final editorial decisions rest with human editors.

Human Oversight & Editorial Authority

In compliance with ICD 505 (Analytic Outreach), BLUFify maintains explicit human oversight of all AI-assisted analysis:

  • What the AI does: Claude AI (Anthropic) summarizes source articles, identifies themes, assigns confidence levels, and drafts BLUF assessments. All outputs are timestamped and logged.
  • What human review occurs: Every published briefing passes through an automated editorial pipeline enforcing ICD 203 analytic standards, readability checks, and consistency verification. The editorial team reviews flagged items before publication.
  • Limits of AI analysis: AI cannot independently verify claims, interview sources, or exercise news judgment in breaking situations. AI outputs are constrained by the quality and breadth of available RSS feeds. Political and source bias is audited algorithmically but not eliminated.
  • How to use BLUFify assessments: BLUFify briefings are a starting point for informed awareness, not a substitute for primary source verification on consequential decisions. Confidence levels and source counts are provided to support your own judgment.
  • Accountability: All AI-assisted analysis is disclosed per ICD 505. Questions or concerns may be directed to [email protected].

This statement is reviewed and updated as the platform evolves. Last updated: March 14, 2026.

Corrections Policy

BLUFify™ is committed to accuracy. When errors occur:

  1. Prompt Correction: Errors are corrected as soon as identified
  2. Transparent Disclosure: All corrections are clearly labeled with date and nature of error
  3. Original Preserved: Strikethrough of original text preserves the record
  4. Public Log: All corrections are logged in our corrections archive

To report an error, contact [email protected]

Data Sources

BLUFify™ aggregates content from 176+ sources including:

  • Wire Services: AP, Reuters, AFP (highest reliability)
  • Major Publications: NYT, WSJ, Washington Post, BBC, Financial Times
  • Broadcast Networks: PBS, NPR, CBS, NBC, ABC, CNN, Fox
  • Think Tanks: CSIS, Brookings, CFR, RAND, Atlantic Council, Heritage, Cato
  • International: Al Jazeera, DW, France 24, Nikkei Asia, Globe and Mail
  • Specialty: Politico, The Hill, STAT News, Ars Technica, Bloomberg

Source selection prioritizes cross-spectrum representation and high factuality ratings. View our complete source directory.

Questions about our methodology?

Contact Us