Transparency & Methodology

How we generate trust scores, where the data comes from, and what independence means to us.

Independence Policy

Swanum does not accept paid placement, sponsored reviews, or vendor payments of any kind. No AI tool company can influence their score. Every analysis is generated from publicly available data and community signals.

Data Sources

  • Reddit — community discussions, bug reports, user complaints via Arctic Shift API
  • Web search — recent news, changelog announcements, incident disclosures via DuckDuckGo
  • Google Trends — search interest over time and week-over-week change
  • VS Code Marketplace — install counts for IDE extension tools
  • Vendor documentation — official privacy policies, security pages, compliance certifications

Trust Score Calculation

Scores range from 0–100 and are generated weekly by a structured AI analysis pipeline (Gemini 2.5 Pro). The model evaluates the following dimensions:

Privacy & Data Handling Data retention, training opt-outs, PII exposure, telemetry disclosure
Security Posture Known CVEs, incident history, responsible disclosure, SOC2/ISO certifications
Community Sentiment Reddit discussion tone, complaint frequency, trust signals from developers
Enterprise Compliance GDPR, HIPAA, SOC2, SSO support, data processing agreements availability
Transparency Changelog quality, incident communication, policy clarity
Trend Signals Search interest direction, community growth or decline, recent news sentiment

Score Interpretation

70+ Proceed — Strong trust posture. Suitable for enterprise evaluation.
55–69 Conditional — Some concerns noted. Review specific risk factors before procurement.
<55 Evaluate Carefully — Significant risks identified. Extended due diligence recommended.

Update Cadence

Reports are generated every Monday at 03:00 UTC. Each run re-scrapes live data so scores reflect the current week's signals, not a snapshot from months ago.

Corrections & Disputes

If you believe a score is factually incorrect (wrong product referenced, outdated CVE data, etc.), please contact us. We investigate all factual disputes within 5 business days. We do not adjust scores based on vendor preference — only verified factual errors.