If traffic drops after a Google update, first establish which pages and measures changed. Check tracking, access and indexing before deciding whether the evidence supports a content or ranking investigation. Update timing alone does not identify the cause.
Google’s September 2026 spam update started on September 24 and completed on October 8. Google recorded completion at 01:00 Pacific Daylight Time. As of this guide’s October 8 publication, there is no completed post-rollout week to assess. The announcement confirms the rollout, not the cause of an individual website’s traffic loss. Google’s incident record.
For a service business, the first decision is what needs attention now. A broken analytics tag, an excluded service page and lower search demand require different responses. More than one can happen together.
This guide helps you assemble evidence, prioritize important pages and decide the next action. The worked example is fictional and shows why one falling total should not produce one sweeping diagnosis. Get the worksheet to record your own evidence as you work through the checks.
Start with the metric and the pages that matter
Write down what fell: Search Console clicks, GA4 sessions, delivered enquiries or qualified enquiries. Keep those measures separate.
Search Console records activity in Google Search. Analytics measures activity on your website. Clicks and sessions should not be expected to match exactly. When comparing them, filter Analytics to Google organic traffic and align the page, country, device and dates. Google’s comparison guide.
Start with the pages a suitable buyer uses: services, relevant industries, case studies and buying guides. Record each page’s role. A large decline on an unrelated article may matter less commercially than a smaller loss on a core service page.
Check actual enquiry receipt separately from the analytics event. Define what qualifies an enquiry before reporting it. Google provides distinct events for generating, qualifying and converting a lead; a form event alone does not establish a qualified opportunity. GA4 event definitions.
Choose a comparison you can explain
Record the property, search type, filters, date range and timezone. Exclude preliminary data from a comparison you describe as complete. Search Console’s latest data can be provisional, and its exports can convert unavailable values into zeros. Preserve that distinction in your worksheet. Performance report guidance.
As a working comparison, use equal periods with the same mix of weekdays. Extend the window when a few clicks or enquiries dominate the result. Add earlier periods and year-over-year context when available. Google suggests reviewing up to 16 months for seasonality and gives three-month comparisons as examples. Traffic-drop guidance.
Google’s recommendation to wait at least a full week after an update completes, then compare that week with a week before rollout, is specifically for core updates. It should not be presented as a prescribed waiting period for this spam update. Core-update guidance.
Log releases, tracking changes, campaigns, holidays and changes to your offer. A comparison during the rollout is useful for investigation, but it cannot establish settled post-update performance.
Fix confirmed faults while investigating the wider decline
Check Manual Actions and Security Issues in Search Console. A reported manual action identifies a specific problem and remediation route. An empty report does not rule out automated ranking or spam effects. Manual Actions guidance.
Inspect affected service URLs alongside an unaffected page. Compare Google’s indexed information with the current live page, including indexability and canonical selection. A successful live test does not prove the page is indexed or will rank. URL Inspection guidance.
An unintended noindex, broken form or confirmed security problem deserves action when established. You do not need to prove an algorithmic cause before repairing it.
| What you observe | What to investigate next |
|---|---|
| GA4 sessions fall; matching GSC clicks are broadly stable | Tracking, consent and reporting changes. Test the intended events and verify actual enquiry receipt. The pattern alone does not prove a tracking fault. |
| Important URLs become excluded | Inspect directives, canonicals, redirects and deployment history. Confirm whether exclusion conflicts with the page’s intended purpose. |
| Impressions and positions fall in a consistent query/page group | Check indexing first, then changed pages, relevance and competing results. Update timing is only one part of the record. |
| Impressions fall while comparable positions are stable | Investigate demand, seasonality and query mix. Stable aggregate position alone is weak evidence. |
| Impressions stay similar but clicks fall | Calculate CTR. Examine the same query, country and device group, then titles, snippets and dated search-result observations. |
| Clicks appear stable but enquiries fall | Test the form and review offer, buyer mix and qualification. Establish whether the business outcome is measured consistently. |
These are investigation routes, not automatic diagnoses. Average position can conceal changes in the queries that produced impressions. A present-day search result cannot show exactly what buyers saw before the decline.
Google Trends can support a demand hypothesis, but it reports normalized relative interest. A low-volume B2B term showing zero does not establish zero demand. Trends methodology.
Review practices before rewriting pages
If the evidence still points toward a search-performance problem, review the affected page group against its buyer purpose and relevant policies. Do industry pages explain different requirements, work and evidence, or just replace industry names?
Google’s scaled-content policy concerns manipulation and little added value, regardless of how content was produced. AI authorship by itself is not a diagnosis. Spam policies.
Document a reason for keeping, improving, consolidating or removing each page. Avoid bulk deletion or disavow work based only on falling traffic or a third-party score. Google says most sites do not need its disavow tool. Disavow guidance.
Google’s spam-update guidance says its systems may take months to recognize compliance. That is not a recovery deadline. A technical repair and a policy review are different tasks.
Worked example: two problems in the same business
Constructed teaching example. All figures, URLs, observations and business details below are invented. This is not a client case, a test of Google’s September update, or evidence of recovery.
Imagine a B2B IT provider at example.com. We compare September 10–21 with September 24–October 5, 2026. Both periods contain 12 days with the same weekday composition. The second is during the rollout.
For this fictional exercise, October 5 is the available-data cutoff. That is not a statement about your property’s reporting delay. GSC uses US desktop, Web text-based search; GA4 uses corresponding Google organic landing-page filters. The example assumes reporting timezones have been reconciled.
Page one: IT support
| Metric | Before → during rollout |
|---|---|
| GSC clicks | 200 → 198 |
| GSC impressions | 4,000 → 4,000 |
| GA4 Google organic sessions | 170 → 84 |
| Delivered forms from this page, all sources | 8 → 7 |
| Manually qualified enquiries, all sources | 3 → 3 |
The fictional change log records a tag deployment on September 28. An invented October 8 test, with the required consent in place, finds that page_view is missing while the form is delivered successfully.
The measurement defect is confirmed. GSC clicks fell 1%, while recorded sessions fell approximately 50.6%. We cannot conclude that the tag explains every missing session. We also cannot call the all-source forms organic conversions: their acquisition source is unknown.
The proposed next action is for the analytics owner to repair the configuration and check that the intended events fire at their expected trigger points without unintended duplicates. That repair and retest are pending in this example. Historical sessions remain incomplete; subsequent reporting needs its own review.
Page two: security audits
| Metric | Before → during rollout |
|---|---|
| GSC clicks | 120 → 30 |
| GSC impressions | 2,400 → 600 |
| GA4 Google organic sessions | 102 → 25 |
| Delivered forms from this page, all sources | 4 → 1 |
Here, the invented deployment record shows an unintended noindex added on September 25. The fictional indexed report shows exclusion; the live inspection confirms the current noindex directive.
This page has a genuine measured search decline as well as an indexing defect. Clicks and impressions both fell 75%. That still does not prove the directive explains the entire decline or that the spam update contributed nothing.
The proposed next action is for the developer to remove the unintended directive and verify the intended access and canonical configuration. The repair is pending in this example. Google’s later indexed state requires a separate check. A deployed repair is not evidence of recovered visibility or enquiries.
The decision record
| Finding | Status | Next decision |
|---|---|---|
| IT-support tracking defect | Confirmed in the fictional evidence | Repair and verify event collection. Leave the amount of historical undercount unknown. |
| Security-audit exclusion and search loss | Both confirmed in the fictional evidence | Repair indexability, then verify Google’s processed state and subsequent performance separately. |
| Spam update caused the losses | Unproven | Retain as a hypothesis. Review later complete periods and specific policy evidence. |
One site, two concurrent problems. Neither an all-site rewrite nor “it is only tracking” follows from these observations.
Use the worksheet to leave a useful handoff
Download the blank CSV, the worked fictional CSV or the Markdown workbook with instructions.
For each affected page, record the observation, evidence, alternative explanations, action, owner and acceptance condition. Mark a finding confirmed, hypothetical or unknown. Keep missing values separate from measured zeros.
Query-level records will not necessarily reconcile to your total. Search Console omits anonymized queries from filtered query data. Do not quietly assign that missing traffic to branded or non-branded demand. Performance filtering guidance.
For the next step, use our important-page indexation playbook or GSC, GA4 and CRM measurement guide, depending on the evidence.
If important service pages are affected, our technical SEO work starts with the evidence and the business role of those pages. Book a free Clarity Call with the affected URLs and dates so we can scope the next investigation. Keep private exports and logs in your team’s controlled records until an appropriate review is arranged.