When Google rankings or organic traffic fall, establish which queries and pages changed before assigning a cause. Then check reporting, access and indexing, content, page experience and explicit notices. These 24 items are an investigation checklist, not a published Google scoring formula.
Two focused guides can help with specific findings below: use the pop-up access and operation checks when a dialog blocks reading or interaction; if you manage a private website network, use the PBN resource and maintenance plan to review resource control, published links and continuing upkeep. Match either investigation to an observed problem rather than treating it as the cause of every decline.
Save the onset date, complete comparison periods, affected URLs and recent changes. For a core update, Google recommends confirming that the rollout has ended and waiting at least a full week before comparing with an appropriate earlier period. That observation window is not a reason to postpone repairing a confirmed outage.
Confirm the data and the affected scope
1. Are the property and periods comparable?
Check that the Search Console property covers the intended protocol, host and path. Compare complete periods of equal length and record filters and aggregation. The query-dimension guidance explains why a query table can differ from the chart; do not compare a directory table directly with a whole-site total.
2. Did clicks, impressions or analytics sessions fall?
Separate clicks, impressions, CTR and average position, then compare organic-search sessions in the analytics tool. Definitions differ. If only analytics sessions fell, investigate tracking delivery, consent and configuration before declaring a ranking loss.
3. Is the change site-wide, template-wide or URL-specific?
Break the data down by page, query, country and device. Keep stable comparison pages as well as affected ones. A change in the site-wide average position can reflect a different query mix; it does not mean every keyword moved by that amount.
4. Could demand, seasonality or reporting be involved?
Use Google’s traffic-drop guide to check demand, seasonality and documented reporting anomalies. Trends website data describes relative interest: a low-volume spike or insufficient data cannot establish the size or disappearance of a market.
Inspect access, crawling and indexing signals
5. Did DNS, certificates or the server fail?
Match the incident time to monitoring and logs, retaining affected hosts and URLs. Ping latency is not a full HTTP or page-load measurement and cannot, alone, determine a CDN fault or a ranking effect.
6. Does the response contain the right page?
Inspect the status, body and redirect chain. Google’s HTTP guidance distinguishes 200, 204, redirects, ordinary 4xx responses and 429/server errors. A 200 does not guarantee indexing; an empty page can be a soft 404, while 429 is treated as an overload-related exception.
7. Did a CDN or firewall change crawler access?
Correlate actual Google requests with edge and origin logs. Identify the service, rule and failed resource. An ordinary browser success or a request with a copied Googlebot User-Agent does not establish the real crawler’s access. Investigate with the Cloudflare request-and-rule checks when applicable.
8. Is robots.txt blocking an intended path?
Read the public robots.txt on the correct host and identify the applicable rules. Blocking crawling does not necessarily remove a URL from search and can prevent Google from reading a page-level noindex or canonical.
9. Did an unintended noindex reach production?
Inspect HTTP headers, initial HTML and rendered output. A template or plugin change may have introduced the rule. Correct only pages meant to be indexed, keeping intentional exclusions for private or irrelevant content.
10. Is the canonical pointing to the intended version?
Compare the declared URL with Google’s chosen canonical in URL Inspection, and read both pages. The live test cannot predict canonical selection. Different products should not all point to a category homepage merely because they share a template.
11. Did a migration lose mappings or create bad redirects?
Map old URLs to the appropriate new destinations. Look for loops, missing content and unrelated homepage redirects. Preserve useful established URLs where possible; a catch-all redirect does not recover a missing answer.
12. Are the sitemap and internal routes useful?
Check the real sitemap, child maps, listed URLs and ordinary page links. A sitemap fetched successfully does not mean every URL was indexed. Important pages need relevant navigable entries, rather than an arbitrary internal-link quota.
Read the content and test the page experience
13. Is important content missing after rendering?
Compare initial and rendered text, specifications, images and links. Check APIs, scripts and session-dependent content. For important information available only after interaction, investigate whether it can actually be fetched rather than assuming that a framework name proves success or failure.
14. Can mobile visitors read and operate the page?
Inspect paragraphs, tables, navigation, buttons and forms at relevant widths. Check clipping, overlays and horizontal overflow. Page-experience guidance concerns the overall experience; a loaded image or an automated visibility flag is not proof that its labels are readable.
15. Are images delaying loading or shifting the layout?
Choose assets for actual display size and clarity. Check responsive candidates, intrinsic dimensions, loading timing and the important first-screen image. Use image-compression and visual checks, rather than requiring every image to be smaller than one universal file-size limit.
16. Which CSS, JavaScript or third-party request is a real bottleneck?
Inspect traces, long tasks, request order and actual usage before changing loading. Code unused in one recording is not automatically safe to remove. Preserve required analytics, customer support and business functions, and verify them after justified changes.
17. Are Core Web Vitals field data or lab observations?
Web Vitals guidance distinguishes real-user measurements from lab tools. The current core metrics are LCP, INP and CLS, assessed at the appropriate 75th percentile. A Lighthouse score or TBT is not field INP and cannot prove that speed alone caused a traffic change.
18. Do titles and result presentation match the page?
Compare title, main heading and visible content using the title-link guidance. A template may have introduced generic titles. Stable impressions with fewer clicks still requires query, device and result-presentation analysis; a more sensational headline is not an established repair.
19. Does the page still answer the actual question?
Read the affected page and representative competing answers. Use the helpful-content questions to inspect accuracy, completeness, practical details and original value. There is no universal word count, rewrite percentage or dwell-time threshold that guarantees recovery.
20. Are sources, specifications and dates current?
Check model, version, price, effective date and source. Make substantive corrections and apply publication-date guidance accurately. Changing the year alone does not update incorrect facts, and an example must not become an invented customer case.
Check explicit notices, resource changes and the repair
21. Is there an actual manual-action or security notice?
Inspect Manual Actions and Security Issues separately, following the actual findings. No manual action does not mean no technical, content or ranking problem. Injected malicious content requires fixing the underlying compromise, not just editing a paragraph.
22. Did links really change, or only a tool’s database?
Open the source and destination and inspect the anchor and current response. Search Console’s Links report is not exhaustive. Guangsuan’s resource work follows its quantity, mixed-role and layered approach; one discovered or lost link does not yield a directly calculable ranking gain or loss.
23. Is structured data accurate and eligible?
Check the type, fields and visible content against the structured-data guidelines. Valid markup does not guarantee a rich result, and a lost rich result does not automatically mean the ordinary page left the index. Do not fabricate reviews, video or other entities to fill fields.
24. Was the actual repair verified under comparable conditions?
Keep before-and-after responses and content, then verify the public page, cache and functionality. Observe later crawling, indexing and performance separately. The core-update guide provides no guaranteed recovery deadline; small routine fluctuations do not justify rewriting the whole site.
Turn observations into a repair record
For each finding, record the URL, observed problem, supporting material, change location, public recheck and next search observation. “The template canonical was corrected” is a deployment result; “Google selected the new representative URL” is a later search-processing result; clicks and qualified enquiries are subsequent outcomes.
If the problem is limited to the brand name, use the brand-search investigation. If another site copied the answer, use the original-content and copyright checks. Guangsuan’s SEO service combines technical work, content and backlink resources, with concrete tasks and verifiable delivery.