Methodology

How to read this report

The report keeps historical Google status, GA4 acquisition/content, Search Console performance, and later public checks separate so association is not mistaken for cause.

Historical Google coverage

Every frozen-sitemap URL was inspected successfully through Google Search Console. This is the authoritative source for the index-status categories.

4,663 URLs · Aug 13–17, 2026

GA4 page performance

Landing sessions describe acquisition; page views and active users describe page-path consumption. They never change or prove Google’s index status.

Requested history + 28/90-day windows

Wave 1 comparison

A frozen not-indexed/indexed sample received limited, non-following public GET requests to test basic transport and indexability signals.

120 requests · Aug 17, 2026

Wave 2 follow-up

Retrospective analysis tested nine hypotheses; 50 later public GETs checked redirect sources, counterpart candidates, and historical hub URLs.

50 requests · Aug 18, 2026

Definitions

Google index status

Indexed by Google

Google reported the URL as indexed.

Found, not yet indexed

Google reported discovering the URL but not indexing it.

Crawled, not indexed

Google reported crawling the URL but not indexing it.

Google reports this URL as unknown

URL Inspection returned the exact successful observation that the URL was unknown to Google.

Redirect detected

Google’s observation reported a page with a redirect.

Not found (404)

Google’s observation reported a hard Not found (404) state.

Conclusion labels

Observation, association, hypothesis, or not supported

Direct observation

A directly reconciled count or current response. It confirms only what was observed at that date.

Association

A repeated relationship in this snapshot. Confounding and reverse causation may remain.

Hypothesis to test

A plausible explanation with explicit competing explanations and a test that could weaken it.

Not supported as primary explanation

The available evidence did not support this as a primary explanation. Future independent evidence could change that assessment.

Interpretation

Historical activity bands

Historical activity observed

More than sparse, stale activity was observed in the earlier monthly page-path data.

Limited, older activity

Only one or two sessions were observed, with the latest activity at least three months old.

No activity observed in returned data

No matching activity appeared in the returned data. This is not proof the page was never visited.

Performance definitions

GA4 and Search Console URL metrics

Landing sessions

GA4 sessions attributed to the first page in the session. Engaged sessions and engagement rate use the same landing-page grain.

Page views and active users

GA4 page-path content signals. Page views include repeats; active users are not additive across URLs or suitable for sitewide summation.

Search performance

Dedicated Search Console page totals provide clicks, impressions, CTR, and impression-weighted average position. Average position is not a fixed rank.

URL-level query terms

The URL explorer shows up to ten non-anonymized returned queries per URL/window, requiring at least one click and five impressions. These bounded lists do not equal page totals.

Search Insights query set

The high-level view uses all returned page/query rows for frozen-sitemap paths, including zero-click rows, before applying privacy checks, explicit review, thresholds, and category caps. The published 617-query set is still not an exhaustive query census.

Reporting windows

GA4: requested history from July 2, 2021 through Aug 17 plus latest 28, latest 90, and previous 90 days; earlier monthly evidence first returned rows in April 2024. Search Console page history: Apr 16, 2025–Aug 16, 2026. Search Insights compares adjacent 90-day windows: Feb 18–May 18 and May 19–Aug 16.

Missing rows

Zero observed counts with no returned source row are shown explicitly as “No returned row.” They are not proof of no demand or no visits.

Limitations

What the evidence cannot establish

  • Google does not expose a causal explanation for discovered or crawled-not-indexed states.
  • “Unknown to Google” is the exact successful observation returned at inspection time—not a claim that Google never encountered the URL.
  • GA4 and Search Console performance may be downstream of indexation and is not causal evidence.
  • Search Console excludes anonymized queries, does not guarantee an exhaustive query fact table, and generally consolidates performance to Google’s selected canonical URL. Query sums therefore do not reconcile to dedicated page totals.
  • Search Insights “multiple pages” counts are lower-bound observations, not proof of cannibalization, duplicate intent, or a preferred canonical.
  • GA4 and Search Console use different attribution, identity, date, and timezone rules; their counts should not be expected to reconcile.
  • Wave 1 and Wave 2 current checks do not reconstruct Google’s historical fetch or follow redirect chains.
  • Wave 2 verified observed target responses and self-canonicals, but did not test targets’ current sitemap membership.
  • GSC referrer examples are partial and historical; the bounded checks did not establish a complete current rendered inbound-link graph.
  • Full-page transport-text similarity includes navigation, footer, and boilerplate. It is not rendered main-content or user-value evidence.
  • The hypothesis analysis is retrospective and exploratory. Controlled treatment-effect language requires pre-specified allocation, outcome, schedule, sufficient sample size, and spillover protection.
Privacy note. The client approved exact per-URL GA4 and Search Console performance plus the explicitly reviewed 617-query Search Insights set in the protected report. Static JSON is downloadable to authorized viewers. Raw Google responses, account/property details, inspection links, internal paths, request metadata, and hashes remain excluded; the 20-row cleanup cohort remains analytics-free.