How We Analyze UFO & UAP Encounters
Project Profound applies a rigorous, transparent methodology to the analysis of UFO and UAP encounter testimonies. Every video in our corpus undergoes the same standardized pipeline — ensuring comparability, reproducibility, and academic credibility across thousands of accounts.
Overview
Our UFO/UAP vertical analyzes publicly available YouTube testimonies — first-person encounter accounts, researcher presentations, congressional hearings, and documentary content — using a structured AI-assisted pipeline.
We do not evaluate the truth or falsity of any claim. Our role is to systematically extract, classify, and quantify the phenomenological content of each testimony using standardized scales, enabling cross-case comparison at a scale no human research team could achieve alone.
The corpus target is 5,000+ videosacross 100+ channels, with every account receiving identical analytical treatment regardless of the researcher's prior assessment of credibility.
Classification System
Tiers
- Tier 1 — Encounter
- First-person or reported testimonies of direct UAP/NHI encounters. Full phenomenology + CET scoring.
- Tier 2 — Program / Research
- Expert analysis, whistleblower testimony, congressional hearings, program disclosures. Claims extraction + program intelligence.
- Tier 3 — Excluded
- Music, compilations, pure entertainment, or off-topic content. Not analyzed.
Tracks
- Encounter
- Contains a specific contact event (sighting, abduction, CE5, etc.)
- Program
- Discusses government programs, institutional actions, or policy.
- Debunk / Ambiguous
- Skeptical analysis or content where classification is unclear.
Hynek & Vallée Classification
Every Tier 1 encounter is classified using the widely-recognized Hynek Close Encounter system and Vallée event categories. These are applied by AI with calibrated rubrics, not subjective human judgment.
Hynek Scale
Vallée Categories
UAP-CET Triad
The Contact Experience Triad (CET) is our proprietary three-dimensional scoring system, designed to mirror the NDE triad (cvNDE, Greyson Scale, NDE-TI) for cross-domain comparison.
Evidence Strength Scale
Range: 7–28
7 criteria (witness credibility, perceptual clarity, specificity, corroboration, unpredictability, physical effects, temporal precedence) each scored 1–4.
View full scaleContact Depth Scale
Range: 0–32
16 items across 4 categories (Observation, Entity Interaction, Consciousness Alteration, Transcendent Elements) each scored 0–2.
View full scaleContact Transformation Index
Range: 0–60
12 life-change domains (8 shared with NDE-TI + 2 adapted + 2 UAP-specific) each scored 0–5. Cross-comparable with NDE transformation scores.
View full scalePhenomenology Extraction
Each Tier 1 video undergoes deep phenomenological extraction — a structured AI analysis that identifies and categorizes the specific phenomena described in the testimony:
Program Intelligence (Tier 2)
Tier 2 content (researcher presentations, congressional hearings, whistleblower accounts) undergoes a different analysis focused on extracting verifiable claims about programs, people, and institutions:
- Claims Extraction — Specific assertions with attribution, evidence level, and domain classification
- People & Organizations — Named individuals, their roles, affiliations, and credibility context
- Programs & Technologies — Named programs, budgets, technologies, and institutional connections
- Timeline Events — Key dates, events, and milestones referenced in the content
Analysis Pipeline
Transcript Extraction
YouTube captions are extracted, cleaned, and punctuated using AI-assisted restoration.
Classification
Videos are classified by tier (Tier 1: Encounter, Tier 2: Program/Research) and track (encounter, program, debunk, ambiguous).
Contactee Resolution
Named experiencers are extracted, normalized, and linked to canonical profiles with deduplication.
Phenomenology Extraction
AI extracts structured data: Hynek/Vallée classification, entity descriptions, physical effects, consciousness phenomena, evidence types.
UAP-CET Scoring
Three independent AI passes score Evidence Strength, Contact Depth, and Transformation Index using calibrated rubrics.
Program Intelligence
Tier 2 content undergoes claims extraction, identifying people, programs, technologies, and institutional assertions.
Event Matching
Timeline events are fuzzy-matched against canonical UFO/UAP events (Roswell, Phoenix Lights, etc.) for cross-video intelligence discovery.
Embedding & Indexing
Transcripts are chunked into passages, embedded via OpenAI text-embedding-3-small, and stored for semantic search.
Credibility Scoring
Beyond encounter-level analysis, Project Profound also scores the source credibilityof every YouTube channel in the corpus. This assessment evaluates the channel's production quality, research rigor, engagement practices, and transparency to produce a comprehensive Credibility Score.
Credibility scoring helps researchers distinguish high-quality, well-sourced channels from sensationalized or low-effort content — enabling weighted analysis where higher-credibility sources carry greater analytical weight in aggregate findings.
View our full credibility scoring rubric