API response fields

What the API returns

Every response returns a verified fact plus its full provenance record. DeepInquiry only ships facts that pass a source-grouping check — at least two source groups on different websites per fact, with Wikipedia mirrors and aggregators set aside before scoring. See our verification methodology for how the score is graded.

Each fact response carries:

• fact_text — the verified claim, as written
• verification_tier — gold, silver, or bronze, based on the evidence strength
• independence_score — a 0.0–1.0 measure (rounded to one decimal) of how much the remaining source groups differ in website or mapped parent company and in kind of source (the field name is historical; it does not measure independence of evidence; the score is coarse, and a high value does not mean the evidence is independent)
• cluster_count — number of separate source groups supporting the fact (grouped by website, and by parent company for the websites we have mapped)
• sources — URL, publisher, cluster ID, and domain class for every corroborating source
• category, era, region — classification tags
• citogenesis_flags — historical name: reason codes for sources removed before scoring (Wikipedia, mirror or aggregator sources, sources on the same website as the page, malformed links)
• source_notes — information only: weaker copies (for example a Wikipedia page or a news aggregator) that were set aside because sources of another kind exist; the fact stands on those other sources

Sample response

{
  "fact": {
    "id": 31,
    "fact_text": "Tardigrades can survive in the vacuum of space, withstand pressures six times greater than the deepest ocean trenches, and endure radiation doses 1,000 times what would kill a human.",
    "category": "nature",
    "era": "modern",
    "region": "global",
    "verification_tier": "silver",
    "independence_score": 0.6,
    "cluster_count": 3,
    "sources": [
      { "url": "https://www.nasa.gov/...", "publisher": "NASA",   "cluster_id": 1, "domain_class": "primary_source" },
      { "url": "https://www.nature.com/...", "publisher": "Nature", "cluster_id": 2, "domain_class": "peer_reviewed" },
      "..."
    ],
    "citogenesis_flags": [],
    "source_notes": ["wikipedia_source_excluded"]
  }
}

A 500-fact sample of the corpus is published as MAC-deepinquiry/verified-facts-sample-500 on Hugging Face (CC BY-NC 4.0).

Data update frequency

The corpus grows continuously. New facts pass through our ingest and verification pipeline and land in the API roughly every 15 minutes. A published fact can later be corrected, or have its sources replaced, when a further check finds a better primary source or a wording problem; the API always returns the current version.

What DeepInquiry is not

• Not a search API against the open web. DeepInquiry ships a curated, pre-verified corpus.
• Not a fact-check-on-demand service. Every fact is verified before it enters the corpus, not on request.
• Not a single-source aggregator. Published facts have at least two source groups on different websites.