Scoring and AI review

Deterministic score

Score each component from 0 to 100.

Opportunity =
  0.22 × SEO evidence
+ 0.20 × Local intent
+ 0.20 × Geo keyword value
+ 0.14 × Brand usability
+ 0.12 × Historical fit
+ 0.12 × Portfolio fit
− Risk penalties

Geo keyword value now includes local search volume, CPC, commercial intent, expected qualified leads, rankability, and evidence confidence.

Thresholds

Adjusted score Decision
85–100 Drop-queue candidate
75–84 Manual review
65–74 Watchlist
Below 65 Reject

Confidence-adjust the score:

adjusted_score = raw_score × (0.6 + 0.4 × evidence_confidence)

Risk penalties

  • questionable anchors: −10 to −40
  • repeated topic changes: −10 to −30
  • weak trademark similarity: −15
  • one dominant referring domain: −5 to −20
  • portfolio saturation: −5 to −25
  • modeled keyword demand with weak evidence: −5 to −20

Malware, phishing, strong trademark conflicts, casino/pharma link profiles, or clear deindexing evidence are hard disqualifiers.

AI finalist review

Send only 20–50 structured evidence packets.

Required response:

{
  "decision": "register | review | reject",
  "recommended_use": "money_site | redirect | brand | hold",
  "best_service": "garage floor coatings",
  "best_market": "South Jordan, Utah",
  "best_keyword_clusters": ["garage floor coating", "polyaspartic floor"],
  "estimated_monthly_geo_volume": 320,
  "estimated_monthly_lead_value": 4800,
  "score_adjustment": -4,
  "confidence": 0.88,
  "strengths": [],
  "risks": [],
  "disqualifiers": [],
  "required_human_checks": []
}

Run a second adversarial review on the top five: find the strongest reason not to register each domain. Rank approved domains inside their predicted drop window, not only inside the daily batch.