Verification Pipelines
Overview of all 14 pipelines across three tiers.
Crowdee provides 14 pre-built verification pipelines organized into three tiers based on depth of analysis, involvement of crowd workers, and required inputs. Every pipeline runs on the same underlying infrastructure — the tiers reflect how thorough and resource-intensive the verification is, not separate systems. You choose a pipeline once per run; results include a structured scorecard and an auditable stage-by-stage breakdown regardless of tier.
Pipeline Tiers
Tier 1 — AI-Only
5 pipelines. Fast (~1–3 min). No crowd workers. AI-only analysis of file metadata and content integrity. Best for high-volume screening or situations where a quick signal is sufficient.
Tier 2 — AI + Crowd
6 pipelines. 40–150 min. Include one or more crowd worker or domain-expert review stages after initial AI analysis. Best for cases where consequential decisions depend on the result.
Tier 3 — Context & Multi-file
3 pipelines. 2–90 min. May require no files at all (URL, identity claim) or multiple files used together (cross-modal). Best for URL credibility, identity verification, and corroborating evidence across media types.
Choosing a Pipeline
Use this table to find the right pipeline for your content type and required depth. For borderline cases, start with a Tier 1 pipeline — if the result is inconclusive or the stakes are high, escalate to the corresponding Tier 2 pipeline on the same files.
| Content type | Fast check (Tier 1) | Deep check (Tier 2) | Special (Tier 3) |
|---|---|---|---|
| Image | verify-image-metadata | verify-image-deep | verify-multimedia-cross-modal |
| Audio | verify-audio-technical | verify-audio-deepfake | — |
| Video | verify-video-technical | verify-video-full | — |
| Text / Document | verify-text-coherence / verify-document-provenance | — | — |
| News article | — | verify-news-article | — |
| Social post | — | verify-social-post | — |
| Source / origin | — | verify-source-credibility | — |
| URL | — | — | verify-url-credibility |
| Identity claim | — | — | verify-identity-claim |
| Multiple media files | — | — | verify-multimedia-cross-modal |
Tier 1 — AI-Only
Five pipelines for fast, automated analysis. No crowd workers are involved. These are suitable for triage, bulk screening, or cases where a human-validated verdict is not required.
verify-image-metadata
EXIF extraction, format analysis, and provenance signals for images. Results in ~1–3 min.
verify-audio-technical
Spectral analysis and prosody evaluation to detect audio splicing or synthesis. Results in ~1–3 min.
verify-video-technical
Container metadata, frame-level consistency, and audio-visual sync checks for video files. Results in ~1–3 min.
verify-text-coherence
Claim extraction, internal consistency analysis, and authorship pattern review for text and PDFs. Results in ~1–3 min.
verify-document-provenance
PDF metadata forensics, content layout analysis, and digital signature review. Results in ~1–3 min.
Tier 2 — AI + Crowd
Six pipelines that combine AI analysis with structured crowd worker review. After the AI stages complete, the run enters an awaiting_crowd state while vetted workers and domain experts provide human judgment. Consensus thresholds gate each crowd stage before the pipeline proceeds to the final verdict.
verify-image-deep
Full image authentication: AI metadata and pixel analysis followed by crowd visual review and expert sign-off.
verify-video-full
Comprehensive video deepfake detection combining technical AI analysis with crowd visual review and expert authentication.
verify-audio-deepfake
AI spectral analysis plus native-speaker and audio-expert crowd review to detect synthetic or spliced speech.
verify-news-article
Claim extraction, source credibility scoring, and professional fact-checker crowd review for news and editorial content.
verify-social-post
Context analysis, platform metadata review, and crowd researcher verification for social media posts.
verify-source-credibility
Live web research to trace content to its source, AI synthesis of a source pointer and category, and crowd fact-checker confirmation.
Tier 3 — Context & Multi-file
Three specialised pipelines for scenarios that do not fit the single-file model. Two pipelines accept no files at all and work entirely from context key-value pairs; one pipeline operates on multiple files simultaneously to assess cross-modal consistency.
verify-url-credibility
Domain reputation, content category, and source history assessment for URLs — no file upload required.
verify-identity-claim
Structured identity claim verification using provided context fields and AI cross-referencing — no file upload required.
verify-multimedia-cross-modal
Corroboration analysis across multiple files (e.g. image + audio + text) to detect cross-modal inconsistencies.
How is this guide?
Pipeline Stages
How AI and crowd stages are sequenced within a verification run.
Image Metadata & Provenance Check
Fast AI analysis of image EXIF data, format integrity, and metadata provenance.