How to Recover from a Google Core Update: A 30-Day Action Plan

Why your rankings dropped in a core update

Your traffic graph just fell off a cliff. Rankings you held for months - maybe years - vanished between one Search Console check and the next. If the timing lines up with a confirmed core update, you're not dealing with a technical bug or a manual penalty. You're dealing with Google recalibrating what it considers quality across billions of pages.

We're enhancing Search so you see more useful information, and fewer results that feel made for search engines. Google, on March 2024 spam & core updates - via amsive.com

A core update recalibrates how Google evaluates content quality. Unlike spam updates that target specific violations like buying links or keyword stuffing, core updates adjust the entire ranking formula. Your site did not get worse. Google decided other content better serves the query. The instinct to panic is real, but panic leads to rash decisions: rewriting entire sites overnight, deleting underperforming pages, or chasing whatever tactical trick topped last week's SEO subreddit. None of that works.

What does work is a structured, evidence-based 30-day plan that addresses the actual signals Google has strengthened. This dispatch lays out that plan in the cadence Glenn Gabe and Marie Haynes have field-tested across dozens of recovery engagements: first diagnose, then audit, then improve and prune in parallel, then harden technical foundations. Recovery typically takes effect at the next algorithmic reassessment, not the same day you make changes.

When to start your recovery work

Wait for the rollout to complete before doing anything substantive. Google's guidance, echoed across recovery practitioners, is to give the update at least a full week after completion before analyzing damage. The December 2025 broad core update took 18 days to fully roll out from December 11 to December 29; the May 2022 update took 15 days; the March 2024 core update ran 45 days. Tremors during the rollout can reverse early movement, so acting mid-rollout is acting on incomplete data.

• Announced: Per the Google Search Status Dashboard, confirm both the start date and the rollout-complete date before measuring impact.
• Diagnosis window opens: One full week after rollout completion - earlier readings include tremor noise.
• Active work window: Days 1 to 30 from the post-rollout start date, structured as five diagnosis days, twelve content days, and six technical days.
• Recovery visibility: At the next algorithmic reassessment, which may be the next confirmed core update or an unconfirmed update in between.

Which sites take the biggest hits

Across 2025's updates, impact rates by vertical diverged sharply. The pattern matches what Glenn Gabe has documented since the original Medic update in August 2018: YMYL and review-heavy sites consistently absorb the highest percentage of negative volatility, while sites with thin or templated content compound their exposure.

Two cautions before you assume your category is your problem. First, not every drop is a quality problem. Glenn Gabe distinguishes three causes of broad-core movement: overall site quality, relevancy adjustments, and intent shifts. Relevancy and intent shifts you cannot directly fix - the user wasn't finding what they needed, and Google reassigned the SERP. Second, gradual declines are different from cliffs. Sudden, catastrophic drops are almost always technical (a misconfigured robots.txt, accidental noindex tags, server outages, or a botched site migration). Algorithm impacts roll in over days or weeks.

Days 1 to 5: Diagnose before you fix

The first five days are pure analysis. Resist the urge to start "fixing" things immediately. Hasty changes during or right after a rollout create noise that masks the actual pattern. The deliverable from this week is a single artifact: an impact spreadsheet that drives every later decision.

1. Rule out the imposters. Cross-reference Search Console clicks against GA4 organic sessions. If GSC clicks are holding steady but GA4 shows a drop, the problem is your tracking, not Google. Check GSC's Page Indexing report for accidental noindex tags, robots.txt changes, or server errors. Navigate to Security & Manual Actions - if anything shows there, you have a manual action requiring a reconsideration request, not a core update.
2. Check for AI Overview cannibalization. If impressions are holding steady but clicks and CTR are falling, Google may still be showing your content but an AI Overview above your listing is absorbing the clicks. That is a separate phenomenon from a core update hit and requires a different response.
3. Confirm the rollout window. Use the Google Search Status Dashboard to verify the core update has finished rolling out, and note both the start and end date. Compare your performance in the week after rollout completion against the week before rollout started. Wait at least one full week after completion before drawing conclusions.
4. Segment losses by structure. In Search Console Performance, segment by search type (Web Search vs Images vs Video vs News), then by subdirectory (/blog/, /products/, /guides/) to spot whether the hit is concentrated or sitewide. Sort affected queries by informational vs transactional intent. Pattern recognition here is the foundation of recovery strategy.
5. Test for relevancy and intent shifts. Manually search Google for each affected query. If 80% of listings are new and answer the query in a new way, this is likely a broad shift in what Google considers a relevant result, not a quality problem on your site - no on-page fix will recover those queries. Build the impact spreadsheet with pre-update clicks, post-update clicks, primary keyword, current position, and position change for every URL that lost meaningful traffic.

Days 6 to 30: The structured recovery plan

The remaining 25 days break into four overlapping workstreams: content quality audit, content improvement on priority pages, pruning and consolidation of weak pages, and technical foundation fixes. The framing matches Glenn Gabe's recurring "kitchen sink" approach to broad core update remediation - quality is rarely one smoking gun, it's a battery of them.

Days 6 to 10: Audit content against the new standard

For every page on your impact spreadsheet, score against five dimensions. Experience signals: does the page contain evidence of first-hand experience - original photos, specific testing results, personal observations? Depth versus competitors: pull up the top three results now occupying your old position and compare word count, topic coverage, media richness, and actionable specificity. Author credibility: does the page carry a named author with verifiable credentials? Freshness: are statistics, tools, recommendations, and external links current - Google detects when content was actually modified, not just date displays. Intent alignment: does the page solve the searcher's problem, or does it dance around it with filler before delivering value? Assign each page P1 (high traffic loss, high revenue impact), P2 (moderate), or P3 (limited recovery potential).

Days 11 to 18: Execute improvements on P1 pages

Adding word count without adding value accomplishes nothing. For each P1 page, lead with the answer (move the most useful information to the top), add original data or first-hand evidence (case studies with specific numbers, original screenshots, before-and-after comparisons), create actionable elements (templates, checklists, decision frameworks, calculators), and refresh stale data (statistics, broken external links, references to discontinued tools). E-E-A-T infrastructure is part of this same sprint, not a separate checkbox: add named author bios with LinkedIn URLs and credentials, build an editorial policy page explaining your fact-checking and correction process, bolster the About page with verifiable company details, and stop publishing under generic "Staff" or "Editorial Team" bylines.

Days 11 to 18 (parallel): Prune and consolidate weak content

Sort every page on your site into four buckets. Keep: strong traffic, good engagement, clearly valuable - leave alone or improve lightly. Improve: topically relevant, some traffic, but thin or outdated - refresh candidates. Consolidate: multiple pages targeting the same keyword - merge the best elements into a single authoritative page and 301 redirect the retired URLs. Remove: zero-traffic pages with no backlinks that serve no user or business purpose - use 410 status codes to tell search engines the resource is permanently removed. Recovery work has documented cases requiring removal of 20-25% of total content. Work in controlled batches of five to ten pages with validation between rounds, and update internal links so they point directly to surviving URLs rather than through redirects.

Days 19 to 24: Fix the technical foundation

Content quality carries more weight than technical performance in core updates, but technical failures amplify content weaknesses. Core Web Vitals act as a tiebreaker when content is otherwise comparable. The three metrics for 2026: LCP under 2.5 seconds (preload the LCP image with fetchpriority="high", convert images to WebP or AVIF, eliminate render-blocking CSS and JavaScript), INP under 200 milliseconds (INP replaced FID as the responsiveness metric), and CLS at or below 0.1. Pages with LCP above 3 seconds saw measurably more traffic loss than faster competitors with similar content.

Days 25 to 30: Validate and prepare for the next reassessment

Verify every change is indexed - submit updated URLs for inspection in GSC, validate canonical tags, check that consolidated pages are crawling cleanly, and confirm 410-ed URLs are dropping from the index. Document what changed and when so the next rollout's signals can be attributed to specific work. Recovery typically appears at the next algorithmic reassessment, not the day you finish. Sites that treat E-E-A-T compliance as a checklist exercise without genuine quality investment rarely achieve durable recovery.

What we're seeing in client recoveries

Note: patterns below are aggregated from audits we've run for clients hit during recent core updates, combined with public case studies from practitioners who specialize in algorithm update recovery. The recurring finding is that the sites that recover have done multiple things at once - not one smoking-gun fix.

Marie Haynes' published 2025 case study on an e-commerce recovery following the December 2025 core update describes the same compound pattern: better understanding of user intent and meeting user needs, higher quality blog posts that demonstrated the brand's real-world experience, improved customer service and brand reputation, refreshed outdated content, and added helpful imagery and graphics. None of those moves alone would have recovered the site. Done together over months, they did.

Counter-pattern we see often: sites that scaled content production with AI and saw it work for a year, then watched a core update compress everything. The algorithm did not ban AI content - it identified content lacking human expertise, which mass-produced AI often demonstrates. The hybrid pattern that survives: AI handles expansion, formatting, and structural clarity, while a human contributor supplies the experiential specifics that can't be fabricated. The audit question for any AI-assisted page is whether a genuine expert would recognize it as containing real insight. If the answer is no, the page needs substantive human contribution.

What we're still watching

Four open questions are shaping how we sequence recovery work for clients heading into the next rollout window.

• AI Overview vs core update attribution: Whether GSC eventually exposes AI Overview impressions and clicks as a distinct surface. Today, AI Overview cannibalization and core-update demotion look similar in aggregate (impressions hold, clicks drop) but require different responses. Better signal separation would dramatically shorten diagnosis time.
• Tremor reversibility: How many sites that drop during the initial rollout are partially reversed by mid-rollout tremors versus locked in until the next reassessment. The December 2025 update had a major tremor on December 20 mid-rollout. Knowing tremor behavior helps decide whether to act mid-rollout or wait.
• YMYL author signal weight: Whether named-author signals with verifiable external footprints (LinkedIn presence, conference appearances, citations) carry more weight than just on-page author bios. The pattern in case studies suggests yes, but the lift is hard to isolate from concurrent content quality work.
• Recovery latency without a next confirmed update: Whether unconfirmed updates in between named core updates can drive partial recoveries for sites that did the work. Some sites have recovered without waiting for the next major rollout, suggesting the ranking systems are reassessing continuously even when no public update is announced.