Guide to recovering from a Google core update

Core-update losses can take months to reverse, and recovery is not guaranteed. Google says improvements can appear between major updates, although some sites may need to wait for another core update. Diagnose the loss before choosing what to change.

First, check whether the drop coincides with a core update on the Search Status Dashboard. Second, identify which pages and queries lost clicks and rule out other causes. Third, improve content and the reader experience against Google’s people-first guidance, especially where you can contribute first-hand evidence. Delete content only when it cannot be salvaged. Fourth, monitor improvements between updates and reassess after the next broad core update.

Four principles shape how to approach recovery.

What a Google core update is and is not

A core update changes how Google assesses content across the web. Google’s documentation describes these changes as broad and says they do not target specific sites or individual pages. The update itself does not establish that you violated a policy. It also does not rule out a separate manual action, security issue or technical failure affecting the site at the same time.

Site-wide signals do count, though. The ranking systems guide says the systems are designed to work at the page level, and that site-wide signals and classifiers also contribute to Google's understanding of pages. Google added that wording on November 19, 2024. The same paragraph warns in both directions. Good site-wide signals don't mean every page ranks well, and poor ones don't sink every page. That implies that, for a site with 400 thin posts next to 40 strong ones, the thin posts can be part of how Google reads the strong ones.

Recovery doesn't always wait for a named update. A December 9, 2025 changelog entry adds a section on smaller core updates. Google's stated reason is to make clear that site owners who improve content can reach higher positions without waiting for the next major core update. Google doesn't announce these smaller updates because they aren't widely noticeable. The same documentation keeps the longer clock in view. Some changes take effect in days, but confirming that a site as a whole produces helpful content can take several months, and a few months with no effect may mean waiting until the next core update.

Why core update recovery matters now

A recovery does not guarantee a return to an earlier peak. In March 2025, Google said, as Search Engine Land reported, that future core updates would not necessarily restore every site’s previous traffic. After the May 2026 update, Glenn Gabe reported that he had not yet seen recoveries among many sites he checked that had dropped that January. That is an observation about those sites at that time, not a recovery rate for all sites. In a September 2024 interview, Danny Sullivan explained that search results had changed to reward other formats, including forum and social content, while noting that some sites could surpass their former traffic.

Core updates since 2023, newest first

Google’s Search Status Dashboard lists 13 broad core updates from March 2023 through May 2026. At the October 30, 2026 check, all 13 had completed, and the later listed ranking updates were spam updates.

The table below runs newest first and gives dashboard start dates and incident-page completion dates in US/Pacific time.

No announced broad core update ran between the June 2025 rollout’s completion on July 17 and the December 2025 rollout’s start on December 11. Google can still push smaller unannounced updates between those dates. The HCU recovery cases below associate rebounds with the August 2024, December 2024 and June 2025 broad core updates; they do not establish that every recovery must happen during an announced rollout.

Diagnose the drop before touching anything

Diagnosis separates real ranking losses from AI Overview click loss and Search Console reporting breaks. Google’s documented process starts by waiting for the rollout to finish, then comparing rollout dates against query and page data in these steps:

• Confirm completion: Check the Search Status Dashboard to confirm the update has finished rolling out, and note its start and end dates.
• Wait a full week: Wait at least a full week after completion, then compare that week against a week before the update started rolling out. We use the week before the start, not before completion, because rollouts have run from 6 days (December 2024) to 45 days (March 2024).
• Size the drop: A slide from position 2 to 4 doesn't warrant drastic action, and Google recommends against changing content that's already performing, while a slide from position 4 to 29 calls for a deeper assessment.
• Split by search type: Analyze Web, Images, Video and News separately, since each can move independently.

Check whether the loss coincides with a core update

Check the drop date against a core-update rollout on the dashboard. A matching date is evidence of timing, not proof of cause. Distinguish core updates from spam, Discover and helpful-content updates, and check for technical failures, security issues or manual actions that may overlap. Gabe’s June 2025 analysis separates earlier volatility from the announced core update that began June 30.

The comparison itself is what Gabe calls a delta report. Export landing pages and queries with clicks, impressions, CTR and position for the post-update period against the prior period. Sort by the largest drops and read the pages at the top of the list. Gabe sorts what you'll find into three kinds of loss:

• A relevancy adjustment means Google decided some pages were never that relevant for queries they used to rank for. Gabe notes that some lost queries were a poor fit for the site. He also says more relevant content, whether improved or newly created, can regain some lost rankings. An intent shift means Google now shows a different type of content for the query, such as forums, video, retailers or a tool, based on what it believes users want. The page can be excellent and still lose because the format changed. Consider whether another format would serve that intent and fit your business; do not assume the query is permanently lost.
• With overall site quality problems, the losses trace to content quality, user experience, aggressive or deceptive ads, aggressive affiliate setups, or technical problems that create quality issues. In our reading of Gabe's taxonomy, quality improvements can address this bucket, while changes in relevance or format may address the other two. He says recovery is unlikely until another broad core update rolls out and could take several.

A site whose losses sit mostly in the first two buckets needs a content strategy change before any quality remediation budget.

Rule out AI Overview click loss and data breaks

Google can change rankings during a core update. AI Overviews and other result features can also change clicks without a matching ranking decline. Stable average position with falling CTR is a prompt to inspect the SERP, not proof that AI Overviews caused the loss. Compare like-for-like queries, pages, devices and countries before attributing the change.

In browsing data from 900 U.S. adults in March 2025, users clicked a traditional result on 8% of visits when an AI summary appeared versus 15% when it didn't, which Google disputes.

Ahrefs compared desktop Search Console CTR across 300,000 keywords from December 2023 to December 2025. For keywords that showed AI Overviews in December 2025, observed position-one CTR fell from 7.3% to 1.6%. The often-cited 58% is a gap against modeled expected CTR, not that raw before-and-after decline. Ahrefs sells SEO tools, and neither its study nor Pew’s measures a core update.

Two historical reporting complications can affect these comparisons. Google’s data anomalies page confirms the first; the second rests on practitioner observations:

• Impressions logging error, May 13, 2025 to April 27, 2026. Google says a logging error prevented accurate impression reporting for that whole window. It affected impressions and the related metrics of CTR and average position, but not clicks. Google didn't correct past data. Mueller confirmed the lack of historical correction on Bluesky in May 2026.
• The num=100 change, September 2025. Practitioners saw desktop impressions fall from around September 10, and Brodie Clark’s roundup attributes a finding that 87.7% of 319 properties lost impressions to Tyler Gargula at LOCOMOTIVE. The mechanism is a community theory, not a Google explanation.

Use clicks as the main outcome when comparing periods affected by the May 13, 2025–April 27, 2026 impressions error. Do not use affected impressions, CTR or average position alone to establish recovery. Outside affected periods, these metrics remain useful diagnostic context. Check the anomaly log for any other issue affecting the report or dates you use.

For current AI visibility, use Search Console’s generative AI performance report, which Google says became available worldwide by August 31, 2026. It reports impressions in AI Overviews and AI Mode. Use it alongside matched query, page, device and country comparisons; it does not isolate the cause of a click loss or supply AI-specific click attribution.

Search Console query and page comparison

Run the comparison at query and page level, because you can't identify which section lost from a site-level total. Pull the post-rollout week and the pre-start week, group by page, then by query, and split pages by URL pattern so each directory or template gets its own line.

Our working rule extends Gabe's taxonomy rather than Google guidance. Suspect quality when losses concentrate in one section with position collapsing from page one to page three. Suspect intent shift when losses spread across hundreds of queries where Google now ranks a different content type in the top spots.

Search Console’s table and export cap at 1,000 rows. For larger analyses, use the Search Analytics API or BigQuery bulk export. The API returns top rows and does not guarantee a complete dataset. Bulk export starts after setup and does not backfill earlier history, so enable it before you need a long comparison.

Why sites lose rankings and why recovery is slow

Core updates change how Google assesses content. Its ranking-systems guide describes both page-level and site-wide signals, without saying that poor signals make every page rank poorly. The core-update guidance says Google may need months to recognize sustained improvement across a site.

Mueller explained on X in May 2024, as reproduced by Search Engine Journal, that a core update can change a ranking formula and thresholds. Formula effects can be ongoing, while threshold changes may need another update. He used thresholds as shorthand for values that require substantial work and data to recalculate. This is a simplified explanation of possible update behavior, not a diagnostic test for any particular site.

Some improvements may affect rankings between updates; other changes may take longer to show. Do not treat the threshold explanation as proof that a particular loss is waiting for a scheduled reset.

Google's documentation gives no number beyond a few days to several months, or the next core update. Its people have gone further.

• At the 2026 Barcelona Search Central event, Gary Illyes discussed recovery timing, according to ROAST’s October 2 attendee recap. Its table labels three to six months as typical for a core-update change and six months to one year as slowest. These are the recap’s labels, not a guaranteed deadline or maximum for every site. Search Engine Journal’s coverage notes that typical is undefined and the underlying sample and measurement period were not reported.
• Marie Haynes’s December 2025 observations describe recovery work taking a year or more, with no guarantee. In her cases, some sites remained suppressed through several core updates.

These are different observations with no shared measurement definition. They should not be compared as equivalent estimates or combined into a recovery deadline.

Remediation: guidance and case observations

The sequence below is our editorial priority, informed by Google’s guidance and the case accounts later in this guide. It is not a measured ranking of interventions by effectiveness or by independent source count. The cases describe bundles of changes, so they do not isolate which action caused a rebound.

Shrink the low-quality indexed footprint

Start by improving pages that serve a real reader. Google’s core-update documentation calls deletion a last resort for content that cannot be salvaged.

It also says that wanting to delete entire sections is likely a sign those sections were made for search engines first, and that deleting the unhelpful content can help the good content on the site perform better. That fits the site-wide signals covered above.

Remove a page when its only reason to exist was a keyword and it can't be made useful. Leave a page alone when it has low traffic but a reader would find it useful. Update a page that's old but still accurate. Google's objection is to pruning by the wrong criteria. Google’s people-first content page, revised in October 2026, treats adding or removing older content primarily to make a site seem fresh as a search-engine-first habit and says it won't help. A Googler, as reported by Search Engine Roundtable, advised focusing on a page's unique value rather than its age or view count.

Four actions cover the footprint, in order of preference:

• Improve: Improvement is the default for any page with a real reader. Google's docs say changes can take effect in days, while the site-as-a-whole judgment takes months.
• Consolidate with a 301: When you've made several thin pages to target one query, merge them and redirect the old URLs to the consolidated page. Google allows that exception in its site-move documentation. It warns that redirecting many URLs to one irrelevant page like the homepage can be treated as a soft 404 and says to keep redirects for at least a year.
• Noindex: Use noindex for pages users need but you don't want indexed (our examples: filtered views, internal tools). The page must remain crawlable so Googlebot can see the rule; do not block it in robots.txt.
• Delete: A 404 or 410 is the right response for pages that can't be salvaged. The Removals tool only hides a URL for about six months.

First-hand evidence in the main content

Put first-hand evidence inside the main content, as the documented recoveries did, though none has been proven to cause a recovery. Google's people-first content page asks whether content clearly demonstrates first-hand expertise, such as having used the product or visited the place. The October 2026 revision added a main content section naming four attributes raters evaluate: effort, originality, talent or skill, and accuracy. The WaterFilterGuru case review reports that the site added real photos of the founder testing products in October 2023. It added test charts and short native videos in April 2024. Go Visit San Diego’s owner described shooting video at hotels. TeachMePersonalFinance’s owner account describes tax-form video tutorials embedded in the articles.

None of the recovery cases reviewed here isolates author boxes or review badges as the cause of recovery. Ready Steady Cut’s owner reported author bylines and profiles among $9,860 of technical work in late 2023, with no uplift twelve months later. The same October 2026 revision cuts the other way. Fabricating creator profiles by pairing AI-generated headshots and invented names and credentials is now named on Google's site-owner page as deception and a signal of a low-quality page. The rater guidelines had carried a similar Lowest-rating rule since at least their January 2025 edition, so bylines added for appearances now carry risk.

UX, ads and getting back in lane

Remove anything that stands between the reader and the main content. Google’s page-experience documentation asks whether a page avoids an excessive amount of ads that distract from or interfere with the main content. Google’s interstitial guidance describes intrusive interstitials and dialogs as page elements that obstruct the user's view of the content, and it recommends banners that take up only a small fraction of the screen. Treat this as hygiene with a modest ceiling. Go Visit San Diego’s owner told Niche Pursuits that he removed ads during remediation but did not think that change helped.

Getting back in lane means cutting or separating sections the site built to chase traffic outside its subject. Our editorial recommendation, untested in the documented cases, is to move any such section that has to exist off the core paths or into its own property.

Core Web Vitals belong in the supporting tier. Google’s page-experience guidance says the vitals are used by its ranking systems and that other page experience aspects don't directly help a site rank higher. The same documentation says Google Search always seeks to show the most relevant content, even when the page experience is sub-par. We'd fix the vitals as hygiene, not as the recovery plan.

What these cases do not establish

The cases reviewed here do not isolate disavow, author boxes, Core Web Vitals or added word count as the cause of recovery. Google’s disavow instructions say most sites will not need the tool. Its conditions concern substantial spammy or artificial links that caused, or are likely to cause, a manual action. A core-update loss alone is not a reason to use it. Mueller has also said disavowing links would not address the core-update loss he was asked about.

No case reviewed here isolates word count or padding as the reason for a rebound. The volume added in Go Visit San Diego's and TeachMePersonalFinance's recoveries was new first-hand content, and no source isolates volume as the cause.

Read recovery rates and cases correctly

Recovery figures depend on their definitions, and the named cases use different metrics. Read cohort-wide rebounds separately from individual recoveries and itemized spending on non-recoveries.

Recovery rates in context

Gabe's cohort figures are easy to misread. In a November 5, 2024 post about roughly 400 sites he'd tracked since the September 2023 helpful content update, he reported that 88 sites, about 22%, had increased 20% or more since the August 2024 core update, and a related report counted 51 sites (13%) that had dropped further. The remaining 65% fall in neither reported bucket, and the post doesn't define them. The base is the depressed post-HCU level, so a 20% rise on a site that lost 90% leaves it at 12% of where it was.

Gabe wrote that many showed a surge then fell back and a few showed full recoveries, calling full recovery an anomaly and noting that most surges were relative, nowhere close to pre-hit levels. He posted before the November 2024 core update began on November 11. The research reviewed here did not establish that cohort’s outcome after the November update. Eight months earlier, in March 2024, Gabe reported no recovery across 380 heavily HCU-hit sites.

A Raptive post last updated September 17, 2025, drawing on more than 4,100 Raptive sites, says nearly 60% of sites hit hard by the September 2023 helpful content update rebounded in June, with weekly traffic up 20% from pre-update levels. Raptive does not name the baseline weeks or define a threshold for being hit hard. It also does not give the size of that affected subgroup; the 4,100-plus sites describe the wider network analysis.

Raptive itself says aggregate conclusions are hard to draw and that results were mixed by vertical, with food, travel and entertainment up and finance and tech down. The reported rebound does not establish a return to pre-September-2023 traffic. Without the baseline dates and subgroup size, it cannot support a general recovery rate.

Named cases with numbers

The owners of the recovered sites made changes well before traffic or rankings rose. Each site stayed at depressed levels for a long stretch before the turn. The first three recovered during a named broad core update. These are owner and third-party accounts with metrics that differ from case to case, and none is Google-confirmed.

• WaterFilterGuru: A July 2026 third-party write-up reports an 83% organic traffic drop within two months of the September 2023 helpful content update and more than a year at 90% or more down. Changes landed around October 21, 2023 and April 14, 2024. The site rose to new traffic highs with the June 2025 core update, about 21 months after the hit. The writer says it's impossible to isolate which change mattered, and no recovered traffic figure is given. The author promotes the traffic database used to find the case, so this is a commercially interested third-party reconstruction rather than owner analytics.
• Go Visit San Diego: The owner told Niche Pursuits in December 2024 that traffic fell about 80% after September 2023. Output went from two or three articles a month to a two-a-day plan starting February 2024, with three or four writers and an editor. Things turned around with the August 2024 update, about 11 months after the hit, and the owner reported in December 2024 that traffic exceeded its previous peak.
• TeachMePersonalFinance: The owner told Niche Pursuits in May 2025 that the site fell from about 14,000 ranking keywords to 16 by summer 2024. He added 80 or more articles of his own, at 12 to 15 a month, on top of about 150 inherited low-quality ones. The December 2024 core update restored rankings and pushed traffic past its previous level, roughly 15 months after the hit. No pruning or technical change is documented.
• HouseFresh: The site’s own post dates the plunge to March 9, inside the March 2024 core update window, and puts the decline since October 2023 at 4,000 Google Search visitors a day down to 200. A ppc.land report from October 2025 says traffic had passed its pre-September-2023 peak, without naming the update. The owner credited hard work and a lot of luck.

The non-recoveries are documented as carefully, with spend itemized:

• Ready Steady Cut: The owner’s October 28, 2024 post lists $3,000 for a helpful content audit, $7,500 for three content audits in four months and $9,860 for technical work between September and November 2023, totalling $20,360 across two agencies. The technical line covered WebP conversion, a schema change, a staging noindex and author bylines. Traffic fell 50% overnight after the September 2023 update and was down 80% by December 2023. The March 2024 core update then locked the loss in. Twelve months on there had been no uplift, and a retrospective quoting the owner’s May 2025 update describes further decline. These are dated accounts, not a current traffic measurement.
• TechRaptor: The owner’s own posts redact the amounts. A ppc.land report from April 2026 itemizes about $20,000 for an Amsive audit in 2024, $10,000 for a Marie Haynes audit before the 2022 hit and $25,000 with SearchMetrics in 2022. That sums to about $55,000, though no source states the total. Traffic went from 2.5 to 3 million a month in 2022 to around 500,000, according to the owner’s 2024 event recap. The research reviewed here found no later report of recovery through April 2026; it did not verify a new traffic measurement for that month.

Where experts disagree

The order of work holds whichever side proves right, but these disagreements decide how much patience and budget to commit.

Improvements can appear between updates, but timing varies

Google’s documentation allows improved content to rise between major updates and warns that broader reassessment can take months. Haynes and Gabe describe cases that remained suppressed through one or several core updates. Mueller’s simplified formula-and-threshold explanation offers context, but it does not tell us which timing applies to a specific site.

Credit for recoveries is unclear

The people closest to the recoveries are cautious about credit. One writer says the changes can't be isolated, and one owner credits luck. Another doubts one of his own changes helped. Gabe reported in June 2025 that sites hit by the September 2023 update surged back across his list, though not all did, after finding in March 2024 that none of his 380 heavily hit sites had recovered. That reads as a systems change as much as a reward for work. These cases do not establish that the changes were necessary or sufficient for recovery. Improve what helps readers, but do not budget as if effort alone determines whether or when traffic returns.

Google's pruning advice points two ways

The footprint rule above sits in the defensible middle of Google's mixed pruning advice. The core updates documentation, last updated December 10, 2025, calls deletion a last resort and steers effort toward improving content, but it also says deleting unhelpful content can help the good content perform better. The people-first page, revised in October 2026, says removing older content just to seem fresh won't help. TechRaptor’s owner described heavy pruning without a reported recovery, while TeachMePersonalFinance’s recovery account documents no pruning. Neither case isolates pruning’s effect.

Small sites compete against platforms either way

HouseFresh’s owners argue the results tilt toward large sites, describing their review buried beneath big-media best-of lists, Quora answers and Google Shopping listings. Google says nothing in a core update targets specific sites, though Sullivan has acknowledged the reward for forum and social content. Either way, a small site competes for the same query against Reddit and retailers, and the page has to be worth choosing over both.

Common recovery mistakes

The mistakes that slow recovery are mostly decisions made before the diagnosis is finished:

• Making speculative changes mid-rollout: Wait for the post-completion comparison before making changes based only on fluctuating rankings. Fix confirmed outages, broken indexing directives or security problems when you find them.
• Rolling back early: Keep improvements in place through the next broad core update.
• Using affected metrics as proof: The historical impressions error compromises related CTR and average-position comparisons for May 13, 2025–April 27, 2026. Use clicks as the main outcome for affected periods, and inspect the anomaly log for other issues.
• Disavowing for a core drop: Google’s conditions concern harmful backlinks that caused, or are likely to cause, a manual action.
• Pruning by age or traffic: Google staff reject both as criteria.

A recovery decision table and where to start

Use broad core updates as review checkpoints alongside continuous monitoring. They are not the only times improvements can appear. Use the table to set the plan and the checkpoints.

Set the pivot threshold before the work starts. Ours is an editorial decision rule, not a Google requirement: if two broad core updates pass after the fixes with no rise in clicks on improved pages, reassess the diagnosis, execution and opportunity cost. For relevance or intent shifts, test whether better-matched content or another format can serve the query and the business. Move effort away from queries that no longer fit, rather than treating all lost queries as permanently gone. Measure AI answer visibility alongside search traffic.

FAQ

How to assess whether a core update contributed to the drop

Check whether the loss coincides with a core-update rollout, then compare the post-completion week with the pre-start week. Timing alone does not prove causation. Rule out other changes and technical problems. Smaller core updates go unannounced, so the absence of a dashboard entry does not exclude algorithm changes.

How long recovery takes

Google gives no fixed deadline: some improvements show in days, while broader reassessment can take months or another update. ROAST’s attendee recap reports three to six months as typical and six months to one year as slowest for a core-update change; those labels are not guarantees. Haynes describes work taking a year or more. The 11-, 15- and 21-month cases in this guide followed the September 2023 helpful content update and should not be treated as a typical range for all core-update losses.

Whether technical fixes alone work

None of the recovery cases reviewed here isolates technical fixes as the cause. Fix confirmed technical problems, but use the diagnosis to decide whether the site also needs quality, relevance or format improvements.

Delete, rewrite, consolidate or redirect

Improve first, consolidate with a 301 second, noindex third and delete last; never choose by age or traffic.

Which content changes help

The documented recoveries included first-hand evidence inside the main content, such as testing photos with test data or video of the place. No case attributes recovery to author boxes alone, and fabricated profiles are now a named low-quality signal.

Whether to disavow links

Most sites do not need to disavow links. Google’s instructions concern substantial spammy or artificial backlinks that caused, or are likely to cause, a manual action and that you cannot get removed. A core-update loss alone is not a reason to use the tool.

Why rankings lag fixes

Mueller’s explanation, preserved by Search Engine Journal, describes ongoing formula effects and thresholds that may need another update. It illustrates why timing can vary; it does not prove that any specific site is waiting for a reset.

---

Running the decision table means stitching Search Console exports, a page inventory with a direction for every URL, a change log tied to update dates and a reporting view a board can read. If maintaining this workflow across separate tools is the bottleneck, book a demo to see how GrowthX runs it in one workspace.