Evidence before rank
Our scoring method shows its working
Discovery finds names. Primary records support narrow claims. Contextual sources surface questions. A rule-based signal keeps those layers separate.
AI-assisted, source-bounded
AI assists with entity deduplication, query expansion and evidence triage. Rule-based scoring reduces paid-ranking influence. It can still miss context or join the wrong entities, so source URLs, access dates, narrow claims, limits and the correction channel remain visible.
The evidence ladder
- Tier 1 — primary: Irish law, regulator or government records tied to a date and scope.
- Tier 2 — corroborating: recognised reporting, court material or an independent authority that supports a narrow fact.
- Tier 3 — contextual: dated reviews, complaints and forums. These raise questions but do not prove them.
- Discovery only: search demand and Affgate inventory. These may prioritise research but never support legal or safety claims.
Signal rules
| Signal | Required basis | What it does not mean |
|---|---|---|
| Green | Current primary Irish record matching exact domain and operator | Guaranteed safety, payout or legality of every product |
| Amber | Open identity, date, transition, scope or complaint evidence | Scam or wrongdoing |
| Red | Official adverse record or corroborated documented adverse evidence | A pile of unverified negative reviews |
Selection contract
The launch queue contains every deduplicated Affgate entity, but only 19 evidence-ready shells are public. Every selected slug is locked for the isolated article pass. The builder wraps one static article file per brand; it does not compose or substitute dossier prose.
Update and correction
Evidence carries an access date. A source changing does not silently change an older capture. Corrections are assessed against the exact domain, operator, claim and date, then recorded through the publication queue.