Methodology — how RTI Wiki sources, verifies, and publishes

Methodology — how RTI Wiki sources, verifies, and publishes — RTI Wiki

Quick Reply: How RTI Wiki sources data, verifies claims, structures articles, and discloses AI assistance. Open methodology for reproducibility.

Open methodology for everything we publish. If a researcher or journalist wants to reproduce or audit any RTI Wiki claim, this page tells them exactly how.

Source hierarchy

We use sources in this order of preference:

  1. Statutes — RTI Act 2005, state RTI rules, related Acts (DPDP, CrPC, CCS Pension Rules, etc.)
  2. Constitutional bodies — Supreme Court, High Court, CIC, SIC orders + judgments
  3. Government circulars — DoPT Master Circular, ministry OMs, CVC manuals
  4. Official portals — sansad.in, rsdoc.nic.in, lsdoc.nic.in, eci.gov.in, prsindia.org datasets, gov ministry portals
  5. Authoritative secondary — PRS India, Common Cause, SNS reports, MyNeta, ADR
  6. News organisations — The Hindu, The Wire, Caravan, LiveLaw, Bar & Bench (specific stories cited)

We do not use sources of unverified provenance (random forums, undated screenshots, anonymous tipoffs).

Data pipelines

Live structured data is mirrored from authoritative APIs:

Dataset Source Update cycle
Rajya Sabha members https://rsdoc.nic.in/MemberGetData/getmemberall Monthly
Lok Sabha members https://sansad.in/api_ls/member Monthly
Parliament bills https://sansad.in/api_rs/legislation/getBills Weekly
MP photos https://sansad.in/getFile/ Quarterly
Wikipedia bios https://en.wikipedia.org/api/rest_v1/page/summary/ Monthly
State portal status Manual periodic verification Quarterly
CIC orders CIC website + manual indexing Continuous

Source URL is preserved for every record. Every CSV download bears a “source” header.

Article structure

Every editorial article follows the 14-block citizen-first format:

  1. Direct-answer block (≤50 words)
  2. Hero image
  3. DPDP banner (cookie compliance)
  4. Lead paragraph
  5. Legal framework (collapsible — citizens skip; lawyers expand)
  6. Key principles (5-7 bullets)
  7. Decision framework (5-7 numbered steps)
  8. Template (copy-ready text in `<code>` block)
  9. Illustrations (subject- or case-wise)
  10. Case-law anchors (2-4 citations with takeaway)
  11. Common mistakes (5-6 bullets)
  12. Pro tips (4-5 bullets)
  13. FAQs (3-5 Q&A)
  14. Conclusion + Related reading + Sources + Last-reviewed date + tags

AI assistance disclosure

We use LLMs (Anthropic Claude, Groq, OpenAI) for:

  • AI Drafter outputs — labelled “AI-drafted”
  • Hindi summary generation — labelled “Beta translation”
  • Outcome Predictor reasoning — labelled “AI-suggested”
  • PIO Reply Checker analysis — labelled “AI-suggested”

We do NOT use LLMs to author the 700+ editorial articles or the case-law summaries without human verification of every fact.

Fact-check protocol

Each article passes through:

  1. Author — drafts + cites
  2. Editor — independent review against sources
  3. Auto-checkerCMS scans for unlinked-citation, broken-link, missing-source
  4. Publication — green Verified badge applied
  5. Re-review cycle — 12 months evergreen, 3 months time-sensitive

When a CIC/SC/HC ruling we cite is reversed or modified, we update the article within 14 days of becoming aware.

Reuse

Editorial content: CC-BY 4.0 with attribution. Datasets: CC-BY 4.0. Tools: free use, no payment, no API rate limit other than fair use.

Reproducibility

Want to reproduce our state-portal status table? CIC backlog stats? Bill counts? Every claim has a Source line at the bottom of its article. Click through. If you find a discrepancy, file a correction.

tag-methodology-transparency-sourcing}}

📱Test our Android app — free beta!Join Beta GroupYou'll receive the install link by email after joining.

Reader signal

Was this article useful?

Tap once if it helped you. These counters show other citizens which pages are worth reading.

- views