CrawlCheck

Lineage

Lineage coverage matrix

NAP claims carry the whole evidence path: a byte locator in the sealed page, the extracted value, the claim it supports, the inputs the rule read, the decision with its span, the points it moved, and the decision leaf in the signed root. This matrix measures how much of that path every finding family has, computed from stored records and nothing else: 236 records, 0 findings, generated 2026-10-08 12:56 UTC.

Records made by the current observer

0 of the sampled records were measured by the Worker version running now (2649041149cf59e3). These cells are exactly what today’s CrawlCheck records; watched sites are re-measured twice a day, so this table refills within a day of every deploy.

Finding familyExact source locatorExtractionClaim / observationRule inputDecisionScore contributionVerifiedGaps
NAP and claims
0 findings on 0 records · 		————————
Robots
0 findings on 0 records · 		————————
Delivery parity
0 findings on 0 records · 		————————
Machine files revenue
0 findings on 0 records · 		————————
JSON-LD and entities
0 findings on 0 records · 		————————
Headers revenue
0 findings on 0 records · 		————————
Capabilities
0 findings on 0 records · 		————————
Mostly code revenue
0 findings on 0 records · 		————————

Every sampled record, including older ones

236 older records were measured by an observer that kept no locators and no spans; their missing cells are a fact about the archive. The daily scan replaces each watched site’s record, so this table converges on the one above.

Finding familyExact source locatorExtractionClaim / observationRule inputDecisionScore contributionVerifiedGaps
NAP and claims
0 findings on 0 records · 		————————
Robots
40 findings on 34 records · ROBOTS_RULES_SHADOWED 13, ROBOTS_NO_SITEMAP 10, ROBOTS_NOT_200 10, RENDER_RESOURCE_BLOCKED 3, ROBOTS_IS_CATCHALL 2, AI_OPTOUT_SET 1, ROBOTS_DISALLOW_ALL 1		97.5%
39/40		100%
40/40		100%
40/40		100%
40/40		100%
40/40		100%
40/40		100%
8/8		locator
why, and the replay (8 of 8 replays reproduced the decision)
  • locator: no locator on this finding (1)
Delivery parity
23 findings on 22 records · ANSWER_ENGINE_REFUSED 16, HOMEPAGE_REFUSED 4, CRAWLER_SERVED_LESS 2, CRAWLER_REDIRECTED_AWAY 1		100%
23/23		73.9%
17/23		100%
23/23		100%
23/23		100%
23/23		100%
23/23		100%
8/8		extraction
why, and the replay (7 of 7 replays reproduced the decision)
  • extraction: no evidence field (6)
  • not replayable 1:
    • no replay is written for HOMEPAGE_REFUSED (1)
Machine files revenue
40 findings on 27 records · SITEMAP_DECLARES_NOTHING 2, MACHINE_CHAIN_HEAVY 16, SITEMAP_NO_LASTMOD 2, NO_LLMS_TXT 18, NO_SITEMAP_FOUND 2		95%
38/40		100%
40/40		100%
40/40		100%
40/40		100%
40/40		100%
40/40		100%
8/8		locator
why, and the replay (6 of 8 replays reproduced the decision)
  • locator: no locator on this finding (2)
JSON-LD and entities
20 findings on 17 records · ENTITY_NO_COORDINATES 16, ENTITY_COLLISION 4		90%
18/20		100%
20/20		100%
20/20		100%
20/20		100%
20/20		100%
20/20		100%
8/8		locator
why, and the replay (7 of 7 replays reproduced the decision)
  • locator: no locator on this finding (2)
  • not replayable 1:
    • the page bytes are not in the store (1)
Headers revenue
40 findings on 34 records · HSTS_MISSING 16, STALE_CACHE_SERVED 21, HSTS_PRELOAD_INELIGIBLE 3		0%
0+40r/40		100%
40/40		100%
40/40		92.5%
37/40		100%
40/40		100%
40/40		100%
8/8		locator, rule_input
why, and the replay (4 of 4 replays reproduced the decision)
  • rule_input: partial: the stored fetches this rule read; it also read the hstspreload.org status lookup, which the manifest does not keep (3)
  • not replayable 4:
    • the rule reads a response header; the sealed manifest keeps bodies by hash, not headers, so this decision replays from the finding's own recorded value only (4)
Capabilities
0 findings on 0 records · 		————————
Mostly code revenue
40 findings on 40 records · PAGE_IS_MOSTLY_CODE 35, CONTENT_NEEDS_JAVASCRIPT 4, PLACEHOLDER_CONTENT 1		87.5%
35/40		100%
40/40		100%
40/40		100%
40/40		100%
40/40		100%
40/40		100%
8/8		locator
why, and the replay (8 of 8 replays reproduced the decision)
  • locator: no locator on this finding (5)

What each column tests

  • Exact source locator: the finding carries a byte range, line or field locator into a fetch the manifest keeps (header locators count as 'recorded', not sealed)
  • Extraction: the finding carries the extracted value(s) it was decided on (evidence)
  • Claim / observation: the finding names the stored fetches it rests on (supported_by with experience ids)
  • Rule input: every input the rule read is in the manifest (exact scope), with the rule's revision recorded
  • Decision: a decision id with a rule span in the trace (parent stage, reads)
  • Score contribution: the score ledger names what this finding did to the grade
  • Verified: the decision leaf folds to the decisions root in the signed manifest; the locator cuts the recorded bytes back out of the sealed file

Each cell counts real findings: yes = the link exists on the record, recorded = it is in the record but not in the sealed manifest (headers), no = missing, with why. A family is complete when no measured cell is missing. The revenue-relevant families are the three carrying the most common findings. A cell measures that a link is PRESENT, not that the finding is right. Replay reruns the rule over sealed bytes where the rule's input is sealed; where it is not (headers, outside probes) it says so instead of passing. JSON: /api/lineage-coverage; one finding’s full path: /api/lineage?id=&code=. Sample: the newest record of every domain in the store (newest 240 plus the portfolio); every finding on each, grouped by family; the verified column and the replay are run on the first 8 findings per family that have a sealed manifest.