Methodology · v2.11

GSGreenwashing Score

Documented DISTINCT greenwashing INCIDENTS at {legal_name} in the last 5 years. EXTRACTION RULES: count ONE EVENT PER UNIQUE INCIDENT, not per news article and not per outlet that republished the same incident. CANONICAL INCIDENT IDENTITY is determined by ONE of: (a) regulator case number / docket ID / agency decision number; (b) court case number; (c) NGO investigation report ID; (d) the canonical date+claim+venue tuple — e.g. a single class action filed on a specific date in a specific federal district court challenging a specific carbon-neutral marketing claim is ONE incident even when five outlets (a sustainability magazine, an ESG-news site, a tier-1 wire, a university blog, a trade journal) republish it on different dates over the following 6 months. When you encounter 5 outlets writing about the same lawsuit / agency action / NGO report, collapse to ONE event — pick the most authoritative primary source URL (court docket > regulator decision > NGO report > tier-1 press > secondary outlet) and put the other URLs in the description field as "also covered by …". Re-publishing on different dates does NOT make distinct incidents; only a NEW underlying action (new docket, new investigation, new NGO report) does. Sources to consult for primary identification: regulator complaints (FTC Green Guides, EU UCPD, AGCM, MHRA-equivalent), NGO call-outs (ClientEarth, Greenpeace, DeSmog), court dockets, retracted ESG claims, fact-checked greenwashing exposés in primary press. Return enumerated_events (each: date+url+severity+description with regulator/NGO/court attribution + outcome status). The `date` MUST be the date of the underlying action (filing date / decision date / report publication date), NOT the date of any secondary outlet that republished it. List ALL DISTINCT incidents or none. Server applies event_count_inverse. If you find 0 incidents after exhaustive search, return events=[] (NEVER absent).