Crawl overview and statistics

Begin with run state and coverage

Overview combines the selected run's published Site Health, issue totals, categories, URL distribution, crawl outcomes, HTTP classes, recent health points, and severity movement. It is an orientation layer, not a root-cause report.

• Confirm that secondary publication finished.
• Compare requested, discovered, admitted, crawled, skipped, failed, and still-waiting scope.
• Check whether summary counts are present and exact enough for the claim.
• Open the issue group or URL set behind the visible pattern.

Read the score with its publication state

Site Health is a 0 to 100 display derived from a versioned basis-point calculation. Four categories are required for publication: crawlability and availability, indexability and canonicalization, content and metadata, and links and architecture. Structured and international, performance and resources, and security and protocol contribute when eligible evidence exists.

Confidence is the minimum of primary coverage, category coverage, and materialization coverage. Below 70 percent confidence the score is not published; 70 to below 85 percent is provisional; 85 percent or more is measured when the other publication rules pass.

Understand what the score weights

Do not compare incompatible health points

A health delta is comparable only when score, ruleset, eligibility, and scope fingerprints match; both runs are measured; materialization is exact; neither evidence set is truncated; and primary coverage differs by no more than five percentage points.

If those conditions fail, compare the underlying issue, URL, and coverage evidence instead of inventing a trend from two colored numbers.

Separate discovered, admitted, and processed URLs

Discovered is everything the run found. Admitted is the bounded set accepted under host, patterns, depth, robots, and URL cap. Crawled, skipped, failed, and waiting describe what happened inside the accepted set. External links and resources have separate secondary counters.

Use Statistics as a reproducible lens

Statistics can scope the whole run or the largest stored directory, segment, locale, host, and template aggregates. It shows response classes, crawl outcomes, click depth, discovery source, indexability, internal-link weakness, structured-data adoption, canonicals, hreflang, response time, and HTML size when those counters exist.

Host and locale scope cards do not pretend to have a matching Page Explorer filter. Directory, segment, and template scopes can hand off their matching filter.

Interpret click depth and link weakness carefully

Click depth is a retained internal-link-path projection from the seed. Start, one, two, three, four to five, six or more, unreachable, and not-linked states are distinct. Not linked is not zero clicks, and deep-page share excludes unknown and unreachable states from its known-path denominator.

Weak internal links currently means at most one incoming internal link. It is a screening condition, not proof that the page lacks authority or should receive a link from every template.

Investigate an unusual depth or link pattern in the page and link evidence, including both ends of the relationship.

A technical state is not automatically a defect

Noindex, canonical-to-other, missing structured data, absent hreflang, redirects, and large HTML can all be intentional. Statistics reports their measured distribution. The issue catalog and page context decide whether a rule was actually triggered.

For markup findings, use structured-data validation to separate a syntax or property issue from rich-result eligibility.

Check Google Index evidence separately from crawl indexability

Google Index shows saved URL inspections from the connected Google Search Console property alongside the selected site's crawl facts. Connect the matching property, inspect the relevant URLs, and check the inspection date and reported last crawl before comparing results.

Indexed, not indexed, and unknown describe the stored inspection verdict. Missing or unknown evidence does not prove that a URL is absent from Google. A successful crawl response does not prove indexing, and this view does not submit indexing requests.

Keep fetch timing separate from user performance

Statistics groups server response time under 0.5 seconds, 0.5 to 1.5 seconds, and above 1.5 seconds, plus downloaded HTML size under 100 KB, 100 to 512 KB, and above 512 KB. These are crawl observations, not Core Web Vitals.

The Performance view owns available CrUX field data, PageSpeed or Lighthouse lab evidence, coverage mode, URL sampling, diagnostics, and retry state.

Missing older counters stay missing

Statistics trends use up to ten recent runs with materialized summaries. Markup and performance panels appear only when their availability markers exist. Older runs that predate a counter are omitted instead of being plotted as a false zero.

Move from a distribution to a testable finding

Example: A deep blog archive

Frequently asked questions