Scrutinly vs Veritier

Scrutinly is a web-based fact-checking service for articles, claims, URLs, social posts, and images. It extracts checkable statements from supplied text and evaluates each against current web sources. URL checking fetches the article directly, avoiding manual copying. Results classify claims as Supported, Unsupported, Mixed, or Satirical and provide links to the sources used. A stacked bar summarizes the proportions of Supported, Unsupported, and Mixed claims in a check. Image analysis is in beta: it estimates signs of AI generation or changed context, then gives a probability score and explanation; the service warns that these evaluations may be unreliable. Results remain in search history, and identical content can reuse cached output without consuming tokens. Security features listed by the site include two-factor authentication, PBKDF2 password hashing, and JWT sessions; searches are private to each account and not sold or shared. The Free plan costs nothing and includes 150,000 tokens. Pro costs $12 a month, while Max costs $49 a month.

Veritier extracts checkable claims from text, URLs, PDFs, and images, then compares them with live web evidence or private references. Results provide a verdict, confidence score, plain-language explanation, and source links. The service is available through a REST API, MCP server over Streamable HTTP, local stdio proxy, and browser interface, with parsing and OCR built into its pipeline. The attest_action endpoint checks named records such as npm, PyPI, CourtListener, SEC EDGAR, and the U.S. Code, blocking actions when records cannot be confirmed. Attestations create Ed25519-signed Claim Credentials that can be checked through JWKS and revoked; credentials remain for 400 days. Requests can use webhooks and signed retries. Free-tier text may be retained and used to improve models, while paid-tier payloads are processed ephemerally and excluded from foundation-model training. The free tier allows 25 verifications monthly; paid plans start at $19.99 a month. Verdicts and confidence scores are probabilistic.