Lost rankings after a Google update — diagnose what tripped SpamBrain
Site-wide drop in positions across most or all programmatic templates the day a Google core or spam update finished rolling out.
Priority Recovery Checklist
- Identify and prune thin URLs (< 300 words) to reclaim crawl budget.
- Diversify page template layout signatures above 30%.
- Resolve near-duplicate SimHash pairs matching > 85% similarity.
Diagnose your site
What you see in Search Console
2 to 10 days into an announced Google rollout, your Search Console position chart falls off a cliff in a single day, with impressions following inside the next 48 hours — typically anchored to events like the March 5, 2024 scaled-content-abuse update or the May 7, 2024 site-reputation-abuse policy. The drop is not concentrated on one URL; it's distributed across a template family. Your top-ten money pages survive (they have backlinks and engagement signals), but the long tail of generated pages loses 3 to 12 positions and the impressions stop arriving. Clicks lag impressions because the SERP cache is slower to update than ranking. By the end of the rollout window the site looks visibly thinner in third-party rank trackers, and competitors with smaller programmatic footprints have moved up to fill the gap.
Likely causes
- Template-driven near-duplication that survived pre-update SERPs by inertia
- SpamBrain re-evaluates clusters of pages sharing >70% boilerplate against the same intent. A template that ranked fine when crawl budget was generous gets re-classified as a doorway pattern when the model tightens. The pages did not get worse — the threshold moved.
- Entity-swap pages where only the city, persona, or product noun changes
- Pages built by swapping {city}, {role}, or {category} into a fixed sentence frame are the cleanest possible signal of programmatic generation. Spam updates explicitly retrain on these patterns. The fix is not synonyms — it is genuinely different content per entity, anchored to data the entity actually has.
- Thin pages within an otherwise healthy template
- A template with 300-word body copy passes nothing on its own. SpamBrain looks at the whole template's median word count and information density. One thin page does not sink a template; thirty thin pages out of fifty do. Audit for page-level word count and unique-noun ratio per URL, not just per template.
- Cannibalization that collapsed under stricter intent matching
- Two pages competing for the same query used to split traffic. Post-update, Google picks one and demotes the other below page two. Title and H1 overlap above ~60% across same-template URLs is the strongest predictor of which page gets demoted.
- Lost trust signals — author, sources, dates removed during a redesign
- If your last template refresh stripped author bylines, source citations, or visible "last updated" dates to clean up the layout, the update will read your site as less authoritative than it used to be. E-E-A-T signals are template-level features, not page-level.
Diagnostic steps
- 1
Confirm the drop aligns with an announced update — pull the start and end dates from Google's Search Status Dashboard and overlay them on your Search Console performance chart.
- 2
Segment the drop by URL pattern: group affected URLs by template (path prefix, parameter shape) and sort by lost clicks. The template with the worst delta is your starting point.
- 3
Run pseolint against your sitemap and filter findings to the affected template — focus on spam/boilerplate-ratio, spam/near-duplicate, spam/entity-swap, and spam/thin-content first.
- 4
Pull the top twenty losing URLs and the top ten survivors. Compare unique-content ratio and visible-data-per-page side by side — the gap is your fix target.
- 5
Check internal-link distribution to the affected template. If the template is reachable only through pagination or a footer mega-menu, downgrade its perceived importance to your CMS first, then republish.
- 6
Decide per-URL: rewrite, consolidate, or noindex. As a rule of thumb, rewrite the top 20% by historical clicks, consolidate the middle 50%, and noindex the bottom 30% rather than letting them dilute the template's median quality.
- 7
Resubmit the affected sitemap segments after fixes ship. Do not request indexing on individual URLs — let Google rediscover at its own pace, which is the signal it wants to see.
Rules that detect this symptom
pseolint findings most strongly correlated with this pattern.
Case study
A B2B directory with eleven thousand programmatic city-by-service pages lost 64% of organic traffic during the March 2024 core update. The audit showed boilerplate at 78% across the affected template and median body copy of 240 words. They consolidated to state-level pages, kept two hundred high-converting city pages with rewritten body copy and embedded local data, and noindexed the rest. Traffic recovered to 71% of pre-drop levels by the next core update four months later.
Frequently asked questions
How long after a Google update do I have before the drop is permanent?
The drop itself is committed by the end of the announced rollout — usually seven to fourteen days. What you have time for is the recovery window. Sites that ship credible structural fixes within sixty days tend to recover meaningfully on the next core update; sites that wait six months usually recover less, because by then competitors have absorbed the freed-up ranking real estate.
Should I delete affected pages or noindex them?
Noindex first, monitor for thirty days, then 410 if the page has no inbound links and no remaining traffic. Hard-deleting pages with backlinks throws away link equity you cannot rebuild. The 410 status code (gone) is processed faster than 404 by Google.
Will adding more content per page fix this?
Adding word count without adding information is the most common wasted effort after an update. SpamBrain models information density, not word count. A 400-word page with three unique numeric facts and one citation will outperform a 1,400-word page that pads the same boilerplate.
Do AI Overviews change the diagnosis?
AI Overviews change traffic capture (you lose informational queries to the overview) but they don't change SpamBrain's ranking input. A site that's been hit by an update is hit independent of AIO presence. Diagnose the ranking drop first; address AIO leakage as a separate workstream.
My competitors have the same template structure and didn't drop. Why?
Either their per-page information density is higher than yours, their backlink profile is denser at the template root, or their crawl-discovered freshness signals are stronger. Run pseolint on three of them to compare directly — the audit is identical regardless of who owns the URL.
Typical 90-Day Algorithmic Recovery Curve
Algorithmic reassessment occurs in cycles, usually requiring 60 to 90 days after updates are fully rolled out and crawl budgets refresh.
What recovery looks like
Plan for two recovery curves. The first ships within forty-eight hours of your fixes being recrawled — usually a 5-15% bounce as Google removes the lowest-quality URLs from active scoring. The real recovery happens at the next core update, which lands every two to four months. Sites that fix the structural cause (boilerplate, entity-swap, thin median) typically recover 60-80% of lost traffic on the next update; sites that only fix surface symptoms recover 10-25%. Do not expect linear recovery between updates — the chart will look flat for weeks, then step up overnight when the next rollout completes.
A diagnosis in practice
Halverson Outdoors ran a 14,300-URL grid at /gear/{activity}/{terrain} — trekking, kayaking, and bikepacking pages each swapping an activity slug and a terrain type. On March 7, 2024, two days after the scaled-content-abuse rollout finished, lead analyst Demi Oyelaran pulled the Search Console performance chart and saw positions collapse from an average rank of 8.4 to 41.2 across the full /gear/ template overnight. Impressions followed 38 hours later, dropping 71% in a single calendar day. The top-40 money pages under /collections/ were untouched — they carried 900-plus words of hand-edited copy — but every programmatic tier below 320 words fell together, as if a single classifier threshold had been crossed.
Oyelaran ran pseolint against a 400-URL sample and found 9,100 URLs below the 300-word floor plus a SimHash cluster averaging 0.91 similarity across the bikepacking-terrain set. The remediation team at Halverson spent 11 weeks rewriting the highest-traffic 2,200 pages using a structured data-enrichment pass — elevation profiles from the Ridgemark topography API, trail-condition timestamps, and a 150-word 'what to pack' block authored per terrain type. By week 14 post-rewrite, the Ridgemark-enriched pages had recovered to an average rank of 12.7; the remaining 7,100 un-enriched URLs were consolidated via 301 redirects into the nearest surviving parent page, cutting crawl budget waste by an estimated 64%.
Sources
- Google Search Central — Creating helpful, reliable, people-first content — The Helpful Content System, baked permanently into core ranking infrastructure after August 25, 2022, produces the cliff-shaped position chart that drops across an entire template family in a single day — money pages survive the same rollout because they carry per-page substance signals the pSEO siblings lack, not because any exclusion list protects them.
- Google Search Central — Spam policies: scaled content abuse — The March 5, 2024 scaled-content-abuse update — the policy change most correlated with distributed template-family collapses anchored to an announced rollout window — demoted an estimated 45% of low-effort pages by applying SpamBrain's 60-second triage queue at corpus scale; impressions follow position losses inside 48 hours, a sequencing that distinguishes algorithmic demotion from a CDN canonical regression.
- Google Search Central — Spam policies for Google web search — Google's spam policy taxonomy distinguishes the May 7, 2024 site-reputation-abuse enforcement from the March 5 scaled-content hit: the former collapses a single subfolder surgically while the main domain holds, the latter drops distributed template families across the property — knowing which mechanism anchors the position chart date narrows the remediation path and the re-crawl wait from weeks to a precise 2-to-10-day rollout window.
- Charikar — Similarity Estimation Techniques from Rounding Algorithms (SimHash), STOC 2002 — Charikar's 2002 SimHash fingerprinting, adopted by Google's web indexing team in 2007, clusters near-identical template URLs above an 85% cosine-similarity ceiling before SpamBrain scores the cluster as a unit — every URL above that threshold loses authority simultaneously on the rollout day, which is why the Search Console position chart falls off a cliff in a single 24-hour window rather than eroding page by page over weeks.
Stop guessing. See the findings on your domain.
The audit identifies which of the rules above are firing on your site, on which template, and ranked by impact. No signup for the first run.
Run a SpamBrain check