Verify a decision

{"actor":"judge","decided_at":"2026-08-27T05:54:37.460Z","decision":"review_revise","investigation_id":"14e3e40f-1c66-42d6-9bfd-64c1a5dad571","new_score":62,"page_slug":"pax-gold-paxg","prev_score":72,"reason":"Computed disputed_pct from claim_findings[] is exactly (3 disputed + 1 unverifiable) / 40 = 0.10, which sits precisely on the boundary between the approve band (0-10%) and the revise band (10-30%); the rules table gives no tie-break for a page landing on the line itself, so this is a judgment call rather than a lookup result. We resolve it toward review_revise for two independent reasons, either of which would settle the case alone: first, at an exact tie we weight toward the more conservative band when the disputed content includes a repeated, directly inverted factual claim rather than a marginal one; second, and decisively regardless of the tie-break, the review flags a high-priority coverage gap (omission of PAXG's allocated-gold/serial-number-lookup feature) which the rules explicitly allow to move a page into review_revise even when disputed_pct alone would not. So the boundary question, while worth stating plainly, does not end up controlling the outcome. On substance: the page states twice (claim_findings[19] and [21]) that PAXG charges a 0.02% on-chain transfer fee and frames this as a disadvantage versus XAUT; the reviewer found this is backwards -- Paxos eliminated that fee in 2024, PAXG currently has zero on-chain transfer fees, and it is XAUT that carries a fee. Notably, the page's own cited source directly contradicts the page's text. This is not treated as severely as a core allegation because it is a technical product-fee detail rather than a claim about fraud, insolvency, or regulatory wrongdoing, but it is repeated and used to draw an unfavorable comparison, so it is not merely cosmetic either. Weighing against this, the page performs unusually well on its most consequential content: it correctly states Paxos's December 2025 conversion to an OCC-regulated national trust (claim_findings[2], [11]), unlike a sibling page reviewed the same day; it never mislabels KPMG's/Withum's attestations as audits (claim_findings[3], [12]); and its NYDFS consent-order section was verified line-by-line against the primary PDF, including exact statutory citations and dollar figures (claim_findings[15]). A timeline date is off by eight months (claim_findings[30], 2018 stablecoin launch) and one citation is dead (claim_findings[18], link_rot), though the page's own source metadata had already flagged that link as unarchivable. One claim (claim_findings[13]) is an unverifiable negative claim spanning years of reports and is treated as a limit of the review method rather than a page defect. Reviewer confidence was 0.8, supporting a firm but moderate correction rather than escalation to denial.","score_delta":-10,"sequence_num":3,"submission_content_hash":null,"submission_id":null,"submission_kind":null,"submission_valence":null,"v":1}