Lizely SEO Analyst Observes Search Console Indexing Report Mislabel as Verified Observation, Not Official Change

Pipeline stage diagnosis

Pipeline stage diagnosis asks which layer of the crawl-to-index pipeline actually moved during the affected window, and the evidence places the fault at the reporting layer rather than discovery, crawl, render, canonicalization, or index selection. Search Console briefly mislabeled indexed pages as "Crawled, currently not indexed" overnight on September 8 into September 9, with the issue clearing on its own within roughly an hour. John Mueller of Google confirmed the glitch on Bluesky the next morning, calling it a "short blip" that the team traced and resolved before most reports went public.

The separation between a reporting error and an indexing error is the core technical distinction: no ranking or indexing behavior actually changed; only the reporting layer briefly showed the wrong status. That makes the failed stage "presentation of index verdict," not any of the upstream stages that would normally warrant a crawl or canonicalization investigation. The verifier surface for a single URL remains the URL Inspection tool in Search Console, which reads Google's own index and returns the current index verdict plus the last crawl date and the canonical Google chose; the site: operator is an incomplete, unranked sample and is not a measurement.

When a dashboard verdict diverges from the URL Inspection verdict or from a live SERP check, the bug is now established as a plausible explanation rather than a deindexing event. com pages showed 171 indexed and 20 not indexed on 10 September 2026, with ten live pages that had been indexed on 1 September moved to crawled, currently not indexed by 8 September and still unindexed when verdicts were re-read on 10 September, holding the drop for at least two more days rather than flickering. Those ten pages returned HTTP 200 by hand, carried no noindex directive, and saw no content change in the window, so a pure "blip" reading does not fit them, and they sit alongside the reporting bug as a separate, still-unexplained pattern that any operator looking at Search Console this week must hold separately from the verified overnight mislabel.

Directive and rendered-output evidence

Directive and rendered-output evidence asks whether fetched and rendered output diverged from declared directives, and on the verified portion of this story the answer is no: declared directives were never the problem. The affected verdicts were a presentation-layer misread of pages that were still indexed and still serving HTTP 200. The reporting glitch surfaced as "Crawled, currently not indexed," but the underlying status was unchanged, which is exactly what makes Mueller's framing precise: nothing breaks if indexing is working as expected and it is just reported differently.

On the older, unverified portion of the story, the same stage is murkier. com pages that moved from indexed on 1 September to crawled, currently not indexed by 8 September and still unindexed on 10 September had no noindex directive, no content change in the window, and the only commits touching them were an edge-cache duration change and a database query optimisation, neither of which alters a single rendered word. Six of the ten were glossary definition pages, against a glossary base of 60 of 191 pages inspected, which is suggestive but not a pattern at a sample of ten.

None of that is a directive or rendering fault by the operator, and none of it is the overnight mislabel either; it is a separately observed drop that does not match the "short blip" mechanism and so cannot be folded into the confirmed incident. The right read is to keep two queues: one for verified reporting-layer misreads, where the fix is to trust URL Inspection and a live SERP check, and one for held drops with no directive or render change, where the question is still open and the verified blip does not answer it.

Cohort experiment, rollback, and recheck

Cohort experiment, rollback, and recheck asks for a control-preserving test, and the right control here is a URL cohort whose index verdict is independently verifiable outside Search Console. The construction that fits this event has three elements. First, define a cohort of pages split into two arms: an exposed arm of pages that showed the "Crawled, currently not indexed" verdict during the affected window, and a control arm of pages on the same property that did not.

Both arms must be checked by URL Inspection and by a live SERP query during the window, since URL Inspection reads Google's own index and returns the current index verdict plus the last crawl date and the canonical Google chose, while a live SERP check confirms what users actually see. com, where ten pages were still unindexed on 10 September after first moving on 8 September. Third, capture the Search Console API verdict, the URL Inspection verdict, and the live SERP verdict on every checkpoint, so a divergence can be attributed to the reporting layer rather than the index.

Rollback does not apply because no change was made; the action is recheck and reconciliation. The experiment also needs a falsifiable shape: if the exposed arm's URL Inspection verdicts and live SERP verdicts agree that pages are indexed while Search Console's aggregated report still says they are not, the verdict is confirmed as a reporting-only fault; if URL Inspection and the SERP both say not indexed, the verdict is a real index decision and is not part of the verified blip. Two pages that appeared to leave the index on 22 August and return on 23 August in the same dataset were trailing-slash legacy URLs that redirect, so their readings reflected the redirect rather than an index decision, which is a useful reminder that any cohort test must filter redirect chains before counting a page as "indexed" or "not indexed" against the cohort.

A natural companion read for operators building this cohort is Reports Observe Google Search Console Page Indexing Report Stale Since June, which documents an older staleness pattern in the same report and sharpens the baseline against which a new cohort reading should be compared.

Knowledge Delta

Knowledge Delta records what is genuinely new, what mechanism the evidence supports, what alternatives remain live, and what would falsify the current read.

New evidence: Search Console briefly mislabeled indexed pages as "Crawled, currently not indexed" overnight on September 8 into September 9, and the issue cleared on its own within roughly an hour, with John Mueller of Google confirming the glitch on Bluesky the next morning as a "short blip" the team traced and resolved before most of the reports went public. Independent third-party data from Matthew Mellinger pulled directly from the Search Console API during the affected window showed up to 20 percent of pages on some smaller sites flipping to the "not indexed" status, concentrated on thinner or older pages, which is the first published API-level confirmation that the mislabel had a measurable footprint and not just a few anecdotal homepages.

Mechanism evidence: The fault sits at the Search Console reporting layer, not in discovery, crawl, render, canonicalization, or index selection. Mueller's framing is that nothing breaks if indexing is working as expected and it is just reported differently, which is consistent with URL Inspection continuing to return the correct index verdict for a single URL while the aggregated Page indexing report briefly drifted.

Alternative explanations that remain live: First, the held drops on rankxai.com, where 10 of 191 pages moved from indexed on 1 September to crawled, currently not indexed by 8 September and were still unindexed on 10 September, with no directive or content change, do not fit the blip mechanism and remain a separate, unverified observation. Second, a pattern of unusual drops reported across the SEO community since the start of April 2026 has not been confirmed by Google, with Mueller writing on 30 April 2026 that he does not see anything exceptional, and Valentin Pletzer asking whether the reports might be a Search Console reporting bug rather than real removal, which means the broader question of whether more pages are being deindexed in 2026 remains open and cannot be settled by this incident.

Decision: Treat this specific incident as a verified reporting-layer mislabel, not a deindexing event, and do not chase phantom deindexing from it; preserve the held drops as a separate queue requiring its own cohort test.

Falsifiable follow-up signal: If a future Search Console mislabel of similar shape occurs, the predicted pattern is that URL Inspection and live SERP verdicts will continue to agree that pages are indexed while the aggregated report disagrees, and that the divergence will resolve within hours rather than holding for days; a held drop that survives past hours is, by this signal, not part of the verified mislabel and must be triaged on its own terms.

Public Action Brief

ACTION LEVEL: Watch only HIGH IMPACT CHANGE: NO WHAT TO DO NOW: Spot-check a small set of homepages and high-value URLs directly in Search Console's URL Inspection tool and in live SERPs during any future window where the aggregated Page indexing report shows a sudden rise in "Crawled, currently not indexed," since URL Inspection reads Google's own index and a live SERP check is independent of the dashboard. Keep a running log of which Search Console and Business Profile reports have been unreliable over the past month so a genuine indexing problem is not lost in the noise the next time a dashboard misfires.

WHAT NOT TO CHANGE YET: Do not reindex, rewrite, or alter directives on pages that flipped to "Crawled, currently not indexed" during the affected window, since the evidence places the fault at the reporting layer and the pages remained indexed. Do not treat the older stale-links report and the missing Business Profile Insights September data as part of the same root cause; Google has not said whether they are connected to the same reporting infrastructure as this indexing status bug. com pages indexed and 20 not indexed on 10 September 2026, against a glossary base of 60 of 191 pages.

MEASUREMENT METRICS: Share of pages in the cohort whose URL Inspection verdict and live SERP verdict agree with the Search Console Page indexing report verdict; count of pages whose verdict divergence persists past hours. com sample); homepages (the segment Mueller singled out as easy to verify independently). com.

WHAT WOULD CHANGE THIS CONCLUSION: A future identical mislabel where URL Inspection and live SERP verdicts also report pages as not indexed, which would move the fault out of the reporting layer; or a held drop lasting past hours that resists URL Inspection and live SERP confirmation, which would require reopening the deindexing question on its own terms. WHEN TO REVIEW: At the next Search Console reporting anomaly, or at the next confirmed Google statement on the older stale-links and Business Profile Insights gaps.

APPLICABILITY: Site owners and SEOs using Search Console's aggregated Page indexing report during the affected window and in any future analogous window. RISK BOUNDARY: This action brief covers the verified reporting-layer mislabel only; it does not authorize action on the unverified held drops or on the broader question of whether more pages are being deindexed in 2026.