Source Research & Credibility Assessment

Live web search and crawling to trace content back to its source, with crowd confirmation of the source's category and credibility.

The Source Research & Credibility Assessment pipeline answers a different question than Crowdee's other credibility pipelines: not "is this URL trustworthy?" but "where did this actually come from?" Give it a claim, a piece of text, or a topic, and it searches and crawls the live web to find the earliest or most authoritative source, identifies any social media accounts associated with it, classifies the source as government, private, or company, and produces a credibility score — confirmed by a fact-checking crowd panel before the final verdict.

Unlike verify-url-credibility and verify-news-article, which reason from an AI model's training knowledge, this pipeline performs live web search and crawling — its findings reflect the current state of the web, not the model's training cutoff.

Pipeline Details

Property	Value
Pipeline ID	`verify-source-credibility`
Tier	2
Estimated Duration	~45 min
Credit Cost	3,200 credits per run
Crowd Stages	Yes (3 `fact_checker` workers, 60% consensus)
Files Required	No files required

Required Inputs

Files

No files required.

Context Keys

Key	Required	Description
`subject`	Required	The content, claim, or topic whose source should be researched.
`known_url`	Optional	A URL already associated with this content. When supplied, the pipeline crawls it directly instead of running a web search.

Stages

Web Research (Web Research)

Searches the live web for the subject (or crawls known_url directly if supplied), then feeds the crawled page content to an AI model to identify the most likely originating source. The stage writes a sourcePointer (URL or description of the origin), sourcePlatformAccounts (any social accounts or bylines found), a sourceCategory (government, private, company, or unknown), and a credibilityScore (0–100) to its artifacts.

Crowd Confirmation (Crowd)

Three workers holding the fact_checker skill tag independently review the AI's source pointer and category against the crawled evidence and confirm or correct it. The stage gates on 60% consensus before proceeding.

Final Scorecard (AI Synthesis)

Combines the web-research findings and crowd confirmation into a final verdict, explicitly noting any disagreement between the AI's source pointer and the crowd's assessment.

Starting a Run

curl -X POST https://api.crowdee.ai/v2/projects/{projectId}/verification-runs \
  -H "X-API-Key: crw_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "pipeline_id": "verify-source-credibility",
    "context": {
      "subject": "Photo claiming to show flooding in downtown Springfield, Nov 2024"
    }
  }'

Example Response

{
  "id": "run_01j9x...",
  "status": "completed",
  "verdict": "authentic",
  "confidence": 78,
  "pipeline_id": "verify-source-credibility",
  "scorecard": [
    {
      "stage": "Web Research",
      "verdict": "authentic",
      "confidence": 74,
      "summary": "Traced to a local news outlet's original report; no earlier source found."
    },
    {
      "stage": "Crowd Confirmation",
      "verdict": "authentic",
      "confidence": 80,
      "summary": "3 of 3 workers confirmed the identified source and category."
    }
  ]
}

For a fast, training-knowledge-only credibility check on a URL you already have, use verify-url-credibility instead — it's cheaper and faster, but doesn't perform live web research.