Web Search Source Allowlist

Web Search Source Allowlist

What it does

This tool manages the per-tenant configuration of trusted source domains for web searches. By restricting which websites the [[tool-web-search-orchestrator]] can query and cite, it reduces the risk of sourcing from unreliable, outdated, or legally irrelevant websites — a primary mechanism for preventing hallucination and citation errors in legal research.

The allowlist is not a wall — it is a priority and filter system. Sources on the list are queried preferentially; sources not on the list are queried only as a last resort and are returned with a lower confidence flag.

Why this matters

Legal research requires source quality standards that general-purpose web search does not enforce. A legal AI that cites a Reddit thread, a Wikipedia article, or a law student blog as authority is dangerous. The allowlist operationalizes professional source standards.

Common failures this prevents:

• Citing an unofficial "explainer" blog for statute text (the statute may have been amended since the blog was written)
• Using a news article as the source for a regulatory requirement (the article may have mischaracterized the regulation)
• Citing a forum post as support for a legal proposition
• Using a law firm client alert as a primary source when the underlying regulation should be cited directly

Default allowlist structure

The default allowlist is organized by source tier and jurisdiction. Tenants can add, remove, or reprioritize entries.

Tier 1 — Official government and courts (always trusted)

Tier 2 — Established legal publishers (trusted secondary sources)

Tier 3 — Bar associations and professional bodies

Tenant-configurable additions

A tenant can add firm-specific trusted sources:

• Their own website (for published legal guides)
• Specific law firm sites they consider authoritative (e.g., Al Tamimi for UAE, Hadef for DIFC, Audi for Lebanon)
• Subscription services the firm has licensed
• Academic institutions specific to their practice areas

Configuration schema

{
  "tenantId": "tenant-xyz",
  "allowlist": {
    "tier1": [
      { "domain": "sama.gov.sa", "jurisdiction": "KSA", "contentTypes": ["regulation", "circular", "press_release"] },
      { "domain": "centralbank.ae", "jurisdiction": "UAE", "contentTypes": ["regulation", "circular"] }
    ],
    "tier2": [
      { "domain": "out-law.com", "jurisdiction": "__multi__", "contentTypes": ["commentary"] }
    ],
    "tenantAdditions": [
      { "domain": "altamimi.com", "jurisdiction": "UAE", "contentTypes": ["client_alert"], "tier": 2 }
    ],
    "blocklist": [
      "reddit.com", "quora.com", "wikipedia.org", "*.blogspot.com"
    ]
  },
  "lastUpdated": "2026-05-01",
  "updatedBy": "admin@firm.com"
}

Blocklist

The following source types are always blocked regardless of tenant configuration:

• Wikipedia (unreliable for legal authority)
• Reddit and other forums
• Anonymous legal Q&A sites (Avvo, JustAnswer, LegalZoom Q&A)
• Law student blogs
• News aggregator snippets (Google News, Yahoo News, Flipboard)
• LinkedIn posts and articles
• Facebook / social media
• Any domain that is not indexed as a legal or government source

Usage by the web search orchestrator

When [[tool-web-search-orchestrator]] executes a query:

1. Primary pass: restrict query to Tier 1 domains (using site: operator) for the relevant jurisdiction
2. Secondary pass: if Tier 1 returns < 3 results, expand to Tier 2 (publishers and bar associations)
3. General pass: if still insufficient, run without site restriction but filter results against the allowlist — only return results from allowed domains
4. Labeling: every returned result is labeled with its tier (sourceType, sourceTier)
5. Blocked domains: any result from the blocklist is silently dropped

Quality signals appended to results

The allowlist system appends quality metadata to every search result:

Related skills

• [[tool-web-search-orchestrator]] — the search tool that reads and enforces this allowlist
• [[tool-legal-data-hunter]] — orchestrator that invokes the web search orchestrator