diff --git a/app/verify/offline.py b/app/verify/offline.py
index e1a79b2..a86e6a6 100644
--- a/app/verify/offline.py
+++ b/app/verify/offline.py
@@ -7,6 +7,10 @@
 * host trust     0..30  — authority of the cited ``source_urls`` (:mod:`hosts`)
 * provenance     0..10  — clean normalized data vs raw-blob-only imports
 
+Categories without domain rules omit consistency and normalize the remaining
+65 points to 0..100. Their subscores retain the original weights, and their band
+cannot be green until domain rules exist.
+
 Hard predicate violations (threads<cores, boost<base, chip postdates device,
 future release) force the band to red regardless of the numeric score.
 """
@@ -111,7 +115,8 @@ def score_record(
     completeness = _completeness(rec.category, data)
     sigs = signals.signals_for(rec.category, data, now_year, soc_release)
     consistency, flags, hard_failed = _consistency(sigs)
-    if rec.category not in RICH_FIELDS:
+    has_domain_rules = rec.category in RICH_FIELDS
+    if not has_domain_rules:
         # Only assess defined structural fields; absent domain rules earn no credit.
         required = getattr(validate, f"{rec.category.upper()}_REQUIRED", {"slug", "name"})
         completeness = W_COMPLETENESS * sum(k in data for k in required) / len(required)
@@ -126,6 +131,11 @@ def score_record(
     provenance = _provenance(data, best_tier)
 
     total = completeness + consistency + host + provenance
+    if not has_domain_rules:
+        # Missing domain rules remove an assessment dimension, not evidence of
+        # validity. Normalize the remaining weights onto the same 0..100 scale.
+        available = W_COMPLETENESS + W_HOST + W_PROVENANCE
+        total = total * 100.0 / available
     subscores = {
         "completeness": round(completeness, 1),
         "consistency": round(consistency, 1),
@@ -135,7 +145,7 @@ def score_record(
 
     if hard_failed:
         band = "red"
-    elif total >= GREEN_MIN and best_tier in (1, 2):
+    elif has_domain_rules and total >= GREEN_MIN and best_tier in (1, 2):
         band = "green"
     elif total < RED_MAX:
         band = "red"
diff --git a/tests/verify/test_offline.py b/tests/verify/test_offline.py
index aab9a00..ceeb862 100644
--- a/tests/verify/test_offline.py
+++ b/tests/verify/test_offline.py
@@ -1,5 +1,8 @@
 """Tier 0 scorer + host classification tests."""
 
+import pytest
+
+from app import validate
 from app.verify import hosts, offline
 from app.verify.common import Record
 
@@ -7,6 +10,52 @@
 NO_SOC: dict[str, str] = {}
 
 
+@pytest.mark.parametrize("category,slug,fields,url,expected", [
+    ("laptop", "acer-chromebook-14-ammok-691",
+     {"ram_gb": 4, "os": "Chrome OS"},
+     "https://huggingface.co/datasets/Ammok/laptop_price_prediction", 53.8),
+    ("monitor", "acer-nitro-monitor-amazonmon-636",
+     {"size_inch": 23.8, "resolution": "1920x1080"},
+     "https://www.kaggle.com/datasets/durjoychandrapaul/amazon-products-sales-monitor-dataset",
+     58.5),
+])
+def test_structurally_valid_variants_are_yellow(category, slug, fields, url, expected):
+    base = "acer-chromebook-14" if category == "laptop" else "acer-nitro-monitor"
+    data = {
+        "slug": slug, "base_model_slug": base, "name": "Example", "brand": "acer",
+        "release_date": "2023-01-01", "verified": False, "source_urls": [url], **fields,
+    }
+    path = f"{category}/acer/2023/{base}/{slug}.json"
+    required = getattr(validate, f"{category.upper()}_REQUIRED")
+    errors = []
+    validate._check_required(path, data, required, errors)
+    validate._check_slug(path, slug, errors)
+    validate._check_source_urls(path, data, errors)
+    validate._check_variant_path(path, data, category, errors, allow_flat=True)
+    assert errors == []
+
+    def score():
+        return offline.score_record(Record(category, path, data), NOW, NO_SOC)
+
+    result = score()
+    assert result.band == "yellow"
+    assert result.score == expected
+    assert result.subscores["consistency"] == 0
+    assert result.flags == ["domain_rules_unavailable"]
+
+    # Strong sources can increase the normalized score, but cannot earn green
+    # without domain consistency rules.
+    data["source_urls"] = ["https://intel.com/example", "https://en.wikipedia.org/wiki/x"]
+    assert score().score >= offline.GREEN_MIN
+    assert score().band == "yellow"
+
+    del data[next(iter(fields))]
+    broken = score()
+    assert broken.score >= offline.RED_MAX  # hard fail, despite a good numeric score
+    assert broken.band == "red"
+    assert "!structural_integrity" in broken.flags
+
+
 def _score(category, data):
     return offline.score_record(Record(category, f"{category}/x.json", data), NOW, NO_SOC)