Guide

SEO Recovery

Last updated 14 July 2026

A quick gander

A rankings drop and a traffic drop are not the same problem. Google Search Console distinguishes three causes: an algorithmic quality reassessment (core update), a spam or scaled content classification, and AI Overviews displacing clicks from pages that still rank. Each has a different diagnosis and a different fix. Before diagnosing any of them, rule out a fourth pattern: two Search Console faults cut reported impressions and average position without touching clicks, so impressions falling while clicks hold steady is a reporting correction rather than a drop. Core update recovery typically requires the next update cycle. Spam recovery requires removing or substantively rebuilding the violating content. AI Overviews displacement is not reversed by improving rankings: the response is to become the source that gets cited.

Interest in “SEO recovery” grew sharply from early 2025 onwards. The timing reflects three overlapping shifts that all produce the same visible symptom (traffic falling) but for different reasons: Google’s expanded use of AI Overviews across more queries, the enforcement of spam policies targeting scaled AI-generated content, and increased update frequency creating more volatility across more sites.

Most recovery guides treat every drop as a core update problem. Many of them are not. Acting on the wrong diagnosis wastes months of effort and can make the underlying problem worse.

How do I know what caused my traffic to drop?

Open Google Search Console and go to the Performance report. Switch to a date comparison: the period before the drop against the period after. Look at three metrics separately.

Impressions fell alongside clicks. The site has lost ranking positions, had pages deindexed, or had content removed from the index. This is an algorithmic, spam, or technical issue. Go to the Pages view, sort by impressions change, and identify which pages were affected and whether there is a pattern: content type, topic cluster, author presence, or page template.

Impressions held but clicks fell. Rankings have not moved significantly, but something is taking the click before users reach your result. From 2024 onwards, the most common cause is AI Overviews appearing for affected queries and answering the question before users click through to organic results.

Impressions fell but clicks held steady. Check the dates before doing anything else. Two Search Console faults produce exactly this shape and neither is a drop in performance. A logging error over-reported impressions from 13 May 2025 to 27 April 2026, and Google has confirmed it will not backfill the correction, so any comparison straddling 27 April 2026 measures corrected data against inflated data.¹ Separately, the removal of the num=100 parameter in September 2025 stripped bot-generated impressions out of the data.² Both events moved impressions, CTR and average position; neither touched clicks. Clicks holding while impressions fall is the signature of the correction, not of lost traffic.

Both impressions and clicks are stable but conversions fell. This is not an SEO problem. The issue is on-site.

Because both faults also distort average position, treat position history that spans them as unreliable for diagnosis. Clicks are the only metric neither event touched, which makes them the trustworthy baseline for any comparison reaching back before May 2026.

Once you know which pattern applies, check the timing against the Google Search Status Dashboard. A drop that coincides exactly with a confirmed core update or spam update points to an algorithmic cause. A sudden sharp drop with no corresponding update suggests a technical issue.

Finally, check the Manual Actions report in Google Search Console. A manual action is an explicit notification that a human reviewer has found a spam policy violation on the site. No notification does not mean no algorithmic issue. Google’s spam detection systems operate continuously without issuing notifications, but a manual action immediately identifies the recovery path.

How do I recover from a core update?

A core update drop means Google’s quality systems have reassessed the site’s content relative to competing pages. No rule was broken. The signal is that competing pages now look more useful, relevant, or trustworthy. Disavowing links, improving Core Web Vitals, or restructuring URLs does not address this; those changes act on signals that core updates do not recalibrate.

The correct response is content quality improvement: pages lacking original insight, first-hand experience, or clear authorship are the consistent underperformers across recent core updates. Recovery typically aligns with the next core update cycle, three to four months later. Changes made between updates are not wasted: they position the site for reassessment. But the improvement alone does not trigger immediate re-ranking.

For a full breakdown of what core updates target and how to respond, see Google core updates.

How do I recover from a spam or scaled content classification?

Spam classifications differ from core update drops in a fundamental way: they are policy violations, not quality reassessments. The recovery path is compliance, not improvement.

Manual actions are the clearer case. They appear in Google Search Console under Manual actions and identify the specific violation. Once the content or links causing the violation are removed or substantively corrected, a reconsideration request can be submitted. Google reviews these manually; turnaround is typically several weeks.

Algorithmic classifications produce no GSC notification. A site may simply stop ranking for affected queries without any explicit signal. The same recovery logic applies: identify the content or pattern that triggered the classification and remove or rebuild it.

Scaled content abuse is the most common modern variant. Google defines it as generating many pages primarily to manipulate search rankings, with little or no value added for users.³ The definition is method-agnostic: thin human-written pages and thin AI-generated pages are treated identically. The question is whether the content adds genuine value, not how it was produced.

The pattern most consistently targeted from 2024 onwards: template-built pages that swap location names, product names, or keyword variants into an identical structure, producing hundreds or thousands of pages where the variable element is most of the page and the fixed structure contributes nothing unique. City × service, year × keyword, location × product. Sites built this way have seen sustained drops across multiple major updates.

A Search Engine Land experiment tracking AI-generated content performance over 16 months found that without demonstrable expertise, authority, or first-hand experience signals, only around 3% of AI-generated pages retained top-100 positions after three months, down from 28% in the first month.⁴ Initial indexing happens readily; sustained ranking does not. The same experiment recorded a partial rebound to 20% in the top 100 after the August 2025 spam update,⁴ so the ceiling is not zero, but the pattern holds: pages of this kind index easily and then decay.

One additional risk: the quality classification for sites with a high proportion of thin programmatic content appears to act at site level. This mirrors how the Helpful Content classifier has operated since it was incorporated into Google’s core systems. Editorial pages on a domain where the majority of content is thin have been reported to lose rankings alongside the low-quality pages.

Recovery from scaled content abuse follows a different arc from core update recovery:

Identify the low-value pages, typically the programmatic templates.
Decide whether they can be made substantively useful. If not, remove them and redirect to the nearest meaningful alternative.
If some pages can be improved, the improvement needs to be real: original data, specific local context, genuine editorial perspective, not rewritten boilerplate.
For manual actions, submit a reconsideration request once changes are complete.
For algorithmic classifications, allow time for Google’s systems to reassess. This typically takes three to six months and does not require a request.

Why has my traffic dropped if my rankings haven’t changed?

If position data in Search Console is stable while clicks have fallen, the most likely cause is AI Overviews appearing for the affected queries. The CTR impact for queries where an AI Overview appears is significant (covered in detail in the AI Overviews article). One caution on the diagnosis: average position is one of the metrics the two impression faults distorted, so confirm the position trend on a window that sits entirely after 27 April 2026 before concluding that rankings held.

The key distinction is between displacement and citation. A site displaced by an AI Overview loses the click. A site cited within an AI Overview gains it. Seer Interactive, tracking 53 brands and 5.47 million queries, found that being cited in an AI Overview delivers 120% more organic clicks per impression than not being cited, and that across 2025 citation was worth two to five times the organic CTR of non-citation even while both were compressing.⁵ Whether the Overview appears is Google’s decision; whether you are cited in it is the part you can influence.

This reframes the response. Getting back to pre-AI Overviews click volumes for affected queries is not a realistic target; the change in how results are presented is structural, not a temporary fluctuation. The more useful question is whether the content is the kind AI Overviews cite or the kind they displace.

Content cited in AI Overviews tends to share characteristics with content that ranks well in organic results: clear structure, direct answers, specific detail rather than generic summaries, and verifiable claims. The difference is granularity. An AI Overview pulls a specific passage, not the page as a whole. A page where each section provides a direct, self-contained answer to a distinct question is better positioned for citation than one structured as a flowing narrative without clear subsections.

Could it be a technical issue?

A sudden sharp drop across many pages simultaneously, with no corresponding update confirmed on the Search Status Dashboard, is more likely a technical issue than an algorithmic one.

Common causes: a noindex tag accidentally applied to live pages, a robots.txt change blocking key directories, broken redirects from a site migration, or a canonical pointing to the wrong URL sitewide.

Unlike algorithmic recovery, technical issues can resolve within days to weeks once corrected. The underlying content has not been penalised; it has been blocked from being seen. The authority already exists. Fixing the blocker allows Google to re-evaluate what was already there.

Check in order:

Google Search Console Page Indexing report (Indexing > Pages): look for a spike in excluded or noindex pages
robots.txt for unintended Disallow rules covering key directories
Site-wide canonical tag configuration
Redirect chains or broken redirects following any recent migration or URL change

For a systematic approach to finding and fixing technical issues, see SEO auditing.

How long does recovery take?

Recovery timelines vary substantially by cause:

Cause	Typical timeline
Technical issue (noindex, robots.txt)	Days to weeks once corrected
Core update quality drop	Next core update cycle: 3–4 months
Manual spam action	Weeks after reconsideration request approved
Algorithmic spam classification	3–6 months after content corrected
AI Overviews click displacement	No return to previous CTR; optimise for citation

Core update recovery depends on how substantive the content improvements are before the next cycle runs. Some sites recover in one cycle; others take two or three. Sites that wait to see whether rankings return before making changes extend the timeline unnecessarily.

AI Overviews displacement does not recover in the conventional sense, though the picture in 2026 is less bleak than the 2025 trend implied. After 18 months of near-continuous decline, organic CTR on Overview-present queries turned upwards in early 2026, climbing from a floor of 1.3% in December 2025 to 2.4% in February 2026. Seer, whose own model had predicted the decline would continue, describes this as a possible levelling out rather than a return to form, and does not expect pre-Overview click-through rates to come back.⁵ Treat that as the realistic ceiling: the format is not going to stop taking clicks, so the response is to pursue citation rather than to try to reverse a ranking drop that did not happen.