Google Manual Actions: 2026 Recovery Guide
The notice never looks like a disaster when it lands. It is one short line inside Google Search Console, filed under Security and Manual Actions, showing a violation type and a button marked Request Review. Traffic usually looks normal for a few more hours. Then the slide begins, and it does not behave like ordinary volatility. Whole sections disappear. Branded queries hold while commercial pages fall off the map. By the end of the week, the category that pays salaries is nowhere on page one.
Here is what nobody tells you about that moment. The technical repair is usually the easy half. The hard half is proving to a person at Google that you found every instance of the problem, not only the instances that were convenient to fix. Most rejected reviews fail on scope, not on sincerity. This guide walks the whole path, from reading the report correctly to writing a request that reviewers actually accept.
What does a realistic recovery actually involve?
A manual action is a human decision about your site, and it is reversible. This guide shows you how to read the Manual Actions report, tell a site-wide match from a partial match, match every violation type to a concrete fix, assemble the evidence a reviewer expects, and write a request that clears review on the first attempt.
Three arguments here run against the usual advice. First, most sites that think they are penalized are not, and the report settles that in about ten seconds. Second, a reconsideration request is a scoping document, not an apology. Reviewers weigh scope far above tone. Third, lifting the penalty restores your eligibility, not your old rankings. Teams that expect a full snap back walk straight into a second crisis.
Scope matters here, so let me set it early. This guide covers human review penalties only. Algorithmic declines from core updates follow a completely different playbook, and confusing the two wastes months.
What is a manual action, and why is it different from a core update hit?
A manual action is a penalty applied by a reviewer at Google after your site breaks a published spam policy. It appears in Search Console with a reason and a scope. A core update decline is algorithmic, silent, and never appears in that report. One is a ruling about rules. The other is a re-scoring of quality.
That distinction decides everything you do next. A human review penalty is lifted through remediation plus a successful request. An algorithmic decline is repaired by making the content genuinely better, which takes months and offers no submit button. If your report says No issues detected, you are in the second situation, and our breakdown of core update recovery is the better starting point.
My honest opinion, after watching this pattern repeat for years, is that the most expensive error in the whole field is treating a ranking slide as a punishment. Teams delete pages that were performing well. They disavow links they legitimately earned. They gut internal linking to satisfy a penalty that was never issued. Check the report first. It costs nothing and takes seconds.
A few reliable tells help. A reviewer penalty tends to be abrupt and section-shaped. An algorithmic move is gradual, query-shaped, and usually lines up with a dated update. Both can be happening at once, which is exactly why you read the report before you build a theory.
Where do you find the Manual Actions report and how do you read it?
Open Google Search Console, find Security and Manual Actions in the left sidebar, then select Manual actions. A clean site shows No issues detected. Anything else lists a reason, an affected scope, and a Request Review button. Read all three before you touch anything, because the scope decides how wide your audit has to go.
Check every property you own, not just the one you open daily. A domain property covers all subdomains and protocols. A URL prefix property does not. European teams often run country subfolders, or separate country subdomains under different properties. The penalty may sit on the one nobody has opened in a year. Search Console email alerts depend on each person's permission level and their own notification settings, so they are easy to miss. Make sure at least two people hold owner access, not limited rights.
Before you change a single file, screenshot the entry with the date visible and archive it. You will want that record when you write the request, and you will want it again in six months when a colleague asks what happened. Then run a full crawl so you have a baseline of the site as it stood. Our technical SEO audit process covers the right sequence, and the free SEO report tool gives you a fast snapshot to attach to your documentation.
What does a site-wide match mean compared with a partial match?
A partial match applies to specific URLs, a directory, or a subdomain. A site-wide match applies to the entire property. One manual action can carry either scope, and that scope tells you where to start searching. Partial does not mean small, because a single directory can hold every page that actually earns money.
Do not read partial as minor. A partial match on a single product directory can remove the only pages that convert. I would honestly rather inherit a site-wide match on a small brochure site than a partial match on the money folder of a store doing serious revenue.
Where Google lists example URLs, treat them as a sample and never as the complete inventory. In my view this is the single biggest reason a second request fails. Reviewers pull examples to illustrate a pattern. Finding the remainder is your job, not theirs.
Which violations trigger manual actions, and how do you fix each one?
The reasons cluster into link problems, content problems, deception problems, and third-party problems. Google publishes the underlying rules in its spam policies and Search Essentials documentation, and that list evolves as new abuse patterns appear. The names change over time, so read your report entry against the current policy page rather than an older article.
Work through the type that matches your report entry, and resist the urge to fix everything at once. Naming the violation correctly matters enormously, because a manual action is lifted on evidence about that specific problem rather than on general improvement.
Unnatural links to your site
This fires when your backlink profile shows patterns Google reads as bought, exchanged, or built at scale. Directory blasts, paid guest posts, and sitewide footer placements are the usual culprits. Remove what you can at the source, disavow the residue, and log every outreach attempt with dates. Do not disavow your entire profile in a panic. For healthier acquisition afterward, our link building and digital PR strategy guide is the honest version, and the backlink data inside these link building tools is what you will use to inventory the profile before you remove anything.
Unnatural links from your site
The mirror problem. Your site sells links. Or it runs sponsored reviews and widget placements that pass ranking signals. The fix is mechanical. Tag paid outbound links with the sponsored or nofollow attribute, or strip them out. Then sweep the archive. The posts at fault are usually three years old and forgotten by everyone. In my experience the affiliate-heavy publishers in the United States and Germany are the ones most likely to be sitting on this without knowing it.
Thin content with little or no added value
This category covers doorway pages, thin affiliate pages, near-duplicate location pages, and syndicated copy with nothing original attached. Short is not the crime. Redundant is. Consolidate overlapping pages, improve the ones with genuine demand, and remove or redirect the rest. Our content refresh and pruning guide lays out the decision tree, and pasting a suspect page into the word counter gives you a quick read on how thin it really is.
Scaled content abuse
Mass-produced pages that exist for search engines rather than readers now occupy their own policy lane. Templates, spun text, and unedited generated drafts all qualify. In the report itself this usually surfaces inside a broader major-spam entry rather than as a row carrying its own name. The remediation is unpleasant but simple. Cut the pages that add nothing, rewrite the ones with real demand, and place a human approval gate in front of publishing. Our roundup of AI tools for SEO content covers where automation genuinely helps and where it quietly hurts.
Pure spam
The most severe entry on the list, and the one with the worst odds. It covers automatically generated gibberish, scraped feeds, and sites built entirely to game rankings. If your report says pure spam, partial remediation will not work. The honest path is usually a rebuild around genuine content, not a cleanup pass.
Cloaking and sneaky redirects
Cloaking means Googlebot sees different content than a person does. Sneaky redirects push users somewhere they never asked to go, frequently on mobile only. Plenty of cases are not deliberate at all. A compromised plugin, a rogue ad tag, or an aggressive consent script can each trigger it. Inspect the page with the URL Inspection tool, compare that against the rendered page in Chrome, and audit every third-party tag. Our redirects and site migration guide explains safe redirect behavior.
Hidden text and keyword stuffing
White text on a white background, zero-opacity blocks, content pushed off screen, and paragraphs of repeated city names buried in the footer. Alt attributes stuffed with phrases count too. Audit templates rather than individual pages, because this defect normally lives inside a global component and multiplies across thousands of URLs. Our on-page SEO checklist makes a useful sweep once cleanup is done.
Structured data issue
Markup describing content the page does not display, review snippets a business writes about itself, or product data attached to pages that sell nothing. Remove the ineligible types, align every property with visible content, and retest using the Rich Results Test. Schema.org offers far more types than any site should actually use. Our schema markup guide explains what is worth keeping.
User-generated spam
Comment spam, junk forum threads, empty profile pages, and dead free subdomains. Google holds the host at fault, which surprises people. Clean the spam. Moderate or close the surface. Tag user links with the ugc or nofollow attribute, and stop indexing thin profile pages. Our user-generated content strategy shows how to keep community value without inheriting the liability.
Site reputation abuse
A strong domain rents space to unrelated third-party content, usually coupons, loans, or casino pages sitting inside a subfolder. Large news publishers in the United States and the United Kingdom have been the most visible targets of this policy. There is no clever workaround. End the arrangement, or keep the section entirely out of Search. Half measures read as evasion to a reviewer, and they cost you a review cycle.
Hacked and scraped content
Injected spam pages and scraped copies of other people's work both breach the rules. Hacked cases often surface in the Security Issues report rather than the manual list, and the two can appear together. Clean the injection, patch the entry point, and confirm nothing remains before you request anything. Our guide to checking whether your site is blacklisted and the blacklist check tool give you the fastest confirmation.
How do you build evidence a reviewer will actually accept?
Build one shared document listing every affected URL, the specific problem, the action taken, the date, and the current status. Reviewers scan quickly. A clean sheet of two thousand rows beats three paragraphs of reassurance every single time. Totals at the top matter most, because they show the reviewer your scope in one glance.
Google Sheets works well. Set link sharing to anyone with the link, then paste that link into your request. Put totals at the top: pages removed, pages rewritten, links removed, links disavowed, outreach emails sent. Add before and after screenshots for the worst cases. Keep the raw crawl export too. You may need it for a second pass.
Spot-check your own claims before you submit anything. Open a random sample of the repaired URLs in bulk with the URL opener and look at them the way a stranger would. In my judgment, far more requests fail on a handful of missed pages than on a weak explanation.
What should a reconsideration request actually say?
Three parts, in order. What the problem was and how it happened. What you changed, with numbers and a link to your documentation. What you put in place so it cannot recur. Keep the whole thing factual and shorter than a page, and lead with the numbers rather than the apology.
Name the violation in your own words. Vague lines about better content quality tell a reviewer nothing. Detail does the work. Here is the shape to aim for, with your own numbers in place of these. 412 location pages came from one template with no unique text. 390 were removed and redirected. 22 were rewritten by a named editor with local prices and photos.
Be straight about history. If a previous agency bought links, say so plainly and without theatrics. Reviewers encounter this constantly and it does not count against you. What genuinely counts against you is a request submitted while the problem is still live. Until the fix is deployed everywhere, the manual action stays exactly where it is.
Skip the pleading. No mention of layoffs, quarterly revenue, or how many years you have been in business. None of it moves a reviewer, and it consumes the space your evidence needs. Submit one request and then wait. Repeat submissions never jump the queue, and they can genuinely slow your case down.
How long does the review take, and what do you do while you wait?
Google says reviews take time, and complicated cases take considerably longer. Link cases are usually slowest, because a reviewer has to assess an entire profile. Plan in weeks rather than days, and never stack submissions. Use the wait to protect revenue through other channels instead of refreshing the report every few hours.
Put the wait to work. Keep revenue alive through email, paid search, social, marketplaces, and direct traffic. Bing and Bing Webmaster Tools run apart from a Google penalty, so that channel keeps working. Mark the date in Google Analytics 4 or your Looker Studio dashboard. Nobody should misread the dip three quarters from now.
What you should not do is panic. Do not noindex the site. Do not delete analytics history. Do not launch a redesign in the middle of a review, because you will destroy your ability to attribute whatever happens next.
What do you do if your reconsideration request is rejected?
Read the response for sample URLs, treat them as proof that your scope was too narrow, then widen the audit before resubmitting. A rejection is a scoping note, not a verdict on your business. Resubmit only after the second pass has covered every folder and subdomain built from the same template.
A rejection almost always means one of three things. You missed cases of the same problem elsewhere. You fixed the symptom and left the cause running, such as deleting spam pages while the script still makes them. Or you asked for review before the changes were live and crawled.
Rerun the audit with a much wider net. If the report named thin pages in one folder, examine every folder built from the same template. If it named links, review the profile of every partner and agency you have ever worked with. Then resubmit with a changelog showing precisely what the second pass uncovered. A rejection is not itself an extra penalty, so the main cost is elapsed time, though the original manual action keeps suppressing traffic the whole way through.
What does recovery look like once the penalty is lifted?
You receive a notice that the issue is resolved. Pages become eligible again as Google recrawls them, which takes days to weeks. Rankings return gradually, and frequently not to the previous level. Give the recrawl weeks rather than hours, and judge the outcome against the pages you kept rather than against the traffic you had before the cleanup.
This is where expectations break teams apart. Removing a manual action removes a suppression. It does not add ranking. If part of your former position came from links you have now disavowed, or pages you have now deleted, that share of the traffic is gone permanently, and pretending otherwise makes the next planning meeting deeply uncomfortable.
Speed up the recrawl instead of refreshing the report every hour. Submit fresh sitemaps. Run URL Inspection on your top pages. Check that internal links still point at the hubs you kept. Our XML sitemaps and indexing guide covers the mechanics. Track recovery by section rather than sitewide, because a partial match recovers inside the folder it hit.
How do you make sure this never happens twice?
Write down the rules for link acquisition, content publishing, and user-generated surfaces. Assign a named owner to each one. Then check the Manual Actions report on a schedule instead of after a scare. Owners matter more than checklists here, because an unowned rule is a rule nobody enforces after the third busy week.
The cheapest manual action is the one you never earn. A quarterly rhythm handles most of the risk. Review new referring domains for paid patterns. Sample recently published pages for genuine added value. Purge spam from community surfaces. Re-run a technical crawl. Building real experience, expertise, authority and trust is ultimately what keeps a site out of the gray zone entirely, and no governance checklist substitutes for that.
One rule gets ignored almost everywhere. If you buy a domain, check its age and history with a domain age checker before you build on it. Expired domain abuse is a named policy. Taking on somebody else's penalty is a miserable way to start a project.
Does this play out differently in the United States, Europe, and Korea?
The rules are global. The exposure is not. Different markets fail in different ways, mostly because of how they buy links and how they publish across languages. American sites trip on affiliate scale, European sites on machine-translated catalogs, and Korean sites sit in a market where Google is only part of the picture.
In the United States, the loudest cases involve affiliate scale and rented subfolders on big media domains. In Europe, the common pattern is machine-translated country versions of one catalog. Multiply that across eight languages and it reads as thin duplicate content. Paid link markets are still normal in parts of Germany, France, and Italy. Links inherited from an old agency cause more than their share of trouble there. In South Korea, Google runs alongside Naver, so a penalty takes out a smaller slice of search. That slice is high value, though, and worth defending.
One practical warning for European teams. Consent management scripts that serve stripped or blocked content to crawlers can look remarkably like cloaking. Test what Googlebot actually receives before assuming your setup is safe.
Frequently Asked Questions
How do I know whether I have a penalty or an algorithmic drop?
Open the Manual Actions report in Google Search Console. If it shows No issues detected, no reviewer has penalized you, and your decline is algorithmic. Algorithmic drops are gradual, spread across query types, and usually line up with a dated core update. Human penalties are abrupt and confined to the scope named in the report.
How long does a manual action last?
Until you fix the root problem and pass a review. Manual actions do expire eventually, but Google publishes no fixed term, and the wait can stretch past a year. Waiting is still a bad plan, because an unfixed violation usually earns the action straight back. Most teams spend far longer on the cleanup than on the review itself. Link work often runs several weeks before anything is ready to submit.
Do I need a reconsideration request, or does it clear itself?
You need the request. Google will not revisit your site just because you shipped changes. A manual action can lapse on its own, but that route is slow and unreliable, and the action tends to come back while the violation is still live. Submit the request once the fix is live, crawled, and documented.
Is disavowing links enough on its own?
Rarely. Reviewers want evidence of genuine effort to remove the links first, with the disavow file catching what could not be removed. A disavow file submitted with no outreach record looks like avoidance. Document who you contacted, when, and what happened, then disavow the remainder.
How long should the reconsideration request be?
Shorter than you think. Aim for a few hundred words covering the problem, the fix with numbers, and the prevention plan, plus a link to your evidence document. Long emotional narratives do not help. Reviewers are looking for scope and completeness, not writing quality or apologies.
Can I submit several requests at once?
No. Only one review runs at a time per issue, so a second submission has nothing to act on. Submit once, wait for the outcome, then respond to whatever the reviewer tells you. Patience genuinely is the cheaper strategy here.
Will my rankings fully recover afterward?
Sometimes, but do not plan on it. Lifting the penalty restores eligibility rather than position. If your former rankings depended partly on links you removed or pages you deleted, that portion is gone. Set expectations with stakeholders before the review completes, not after the traffic fails to return.
Does the penalty affect Bing or other search engines?
No. A Google penalty applies to Google's own surfaces, including Search, News, and Discover. Bing, Naver, and other engines run their own systems and their own quality reviews. Still, conduct bad enough to trip one engine often draws notice elsewhere. Treat a clean-up as global, not Google-only.
What happens if I ignore the notice entirely?
The manual action stays in place, and affected pages stay demoted or missing from results. Do not count on it lapsing quietly. Ignoring it makes things worse. The root issue usually keeps making new bad pages while you wait for a fix that never comes.
Can I inherit a penalty when I buy a domain?
Yes, and it happens more often than buyers expect. Penalties attach to the site, not the owner. Before purchasing any expired or aged domain, verify its history, check archived versions of the pages, and where possible get Search Console access from the seller before money changes hands.
Where should you start today?
Open Search Console and check the report. That single step separates the small group with a genuine penalty from the far larger group carrying a quality problem they have misdiagnosed. If the report is clean, close this tab and go read about core update recovery instead. You will save yourself months.
If it is not clean, work in this order. Screenshot and record the entry. Identify the violation type and the scope. Audit far beyond the sample URLs. Repair the mechanism as well as the symptom. Build the evidence document. Then write a short, factual, numbers-first request and submit it exactly once.
Remember the opening scene, the quiet one-line notice that emptied a whole section by Friday. Sites come back from that all the time. They come back through patient, documented work rather than clever arguments. A manual action is really a scoping problem in a scary costume.
My prediction for the next few years is that third-party and scaled-content cases keep growing while classic link penalties shrink, simply because that is where the abuse has moved. Which of those two is your site more exposed to right now?