Google Penalty: What Actually Counts as One, and How to Get It Lifted

A Google penalty is a manual action: someone at Google reviewed the site by hand, decided that pages on it break the spam policies, and applied a specific named action that shows up in Search Console. It is not what happens during a core update, and the report's own documentation does not use the word penalty anywhere. This is written for the person running a SaaS or B2B site who has watched organic traffic fall and now wants to know whether a reviewer was involved at all. It covers the actions Google documents, how to read the report, what a reconsideration request has to contain, and what to do when the report comes back clean.

The short version

• The word "penalty" does not appear anywhere on Google's Manual actions report documentation. The report's vocabulary is "manual action", and the difference is not pedantry: it decides whether you have anything to appeal.
• The two penalty listicles on page one disagree with each other and with Google. One names seven types, the other names twenty, and Google's own list runs to thirty-two entries.
• If the report says no issues detected, you do not have a manual action. The traffic loss is still real, but the cause is somewhere else and so is the fix.
• Google's documentation for the report gives a range, not a deadline: reviews "can take several days or weeks", and link-related requests may take longer. Neither that page nor Google's reconsideration requests page gives a specific number of days.
• Fixing some of the affected pages earns nothing. Google states that a partial fix will not earn a partial return to search results.

What is a Google penalty, exactly?

A manual action, and nothing else. Google's documentation describes it precisely: Google issues a manual action "when a human reviewer at Google has determined that pages on the site are not compliant with Google's spam policies." A person looked, a policy was named, and the site was demoted or removed for that reason.

The same page adds the detail that makes this easy to misread. Most manual actions "will result in pages or sites being ranked lower or omitted from search results without any visual indication to the user." Nothing on your site changes. Nothing in the SERP announces it. The only notification is the report itself and a message in Search Console.

An algorithmic ranking change is a different event with a different cause. Google reassessed the web, your relative position moved, and no reviewer was involved. Nothing appears in the report, nobody explains the decision, and there is no request to file. Our guide to how a core update differs from a manual action covers that boundary in full, and it is not re-opened here.

"Penalty" is industry shorthand. Google's Manual actions report documentation never uses the word, and the policies an action enforces live in Google's spam policies for Google Web Search, which is where the definitions actually sit.

Which manual actions does Google actually document?

Thirty-two, named one by one on a single page. Seventeen of them are labelled as Google News and Discover policy violations: one umbrella entry plus sixteen specific policies covering things like manipulated media, medical content and impersonation. That leaves the fifteen below. Those are the entries Google does not scope to News and Discover, and they are the ones this article covers. Google does not present its list as two groups, so the fifteen-and-seventeen split is our reading of the labels, not a Google heading.

Counted from Google's Manual actions report documentation on 14 September 2026. The labels are Google's, copied verbatim, because the label is what the report will show you.

Manual action	What Google says triggers it
Back button hijacking	The site interferes with browser navigation so a visitor cannot use the back button to return to where they came from.
Site abused with third-party spam	A significant portion of the site is being abused with spam, typically on forums, guestbooks, file uploaders or internal search pages.
User-generated spam	Spam submitted by visitors, usually on forum pages, guestbook pages or user profiles.
Spammy free host	A significant fraction of the sites on a free hosting service are spammy, so the action lands on the service itself.
Structured data issue	Markup outside Google's structured data guidelines: content marked up that is invisible to users, irrelevant or misleading.
Unnatural links to your site	A pattern of unnatural, artificial, deceptive or manipulative links pointing at the site.
Unnatural links from your site	The same pattern in the site's outbound links.
Thin content with little or no added value	Low-quality or shallow pages. Google's own examples: thin affiliate pages, scraped content, low-quality guest posts and doorways.
Cloaking and/or sneaky redirects	The site shows users different pages than Google sees, or sends users somewhere Google did not.
Major spam problems	Aggressive techniques such as scaled content abuse and cloaking, or repeated and egregious spam policy violations.
Cloaked images	Images display differently in Google's results than they do on the site.
Hidden text and/or keyword stuffing	Text hidden from users, or keywords stuffed into a page, in breach of the spam policies.
AMP content mismatch	The AMP version and its canonical page are not substantially the same page.
Sneaky mobile redirects	Mobile visitors are redirected to content that is not available to search engine crawlers.
Site reputation policy	Part of the site carries content inconsistent with Google's site reputation policy. Google adds that this action's impact does not apply to results shown to users in the European Economic Area.

Two entries are the same policy seen from opposite ends. Unnatural links to your site covers links pointing at you; Unnatural links from your site covers links you publish. Different work entirely, and the report names which one you have.

Why do the penalty lists on page one disagree with Google's?

Because they are not copies of it. On 14 September 2026 the result ranking first for this term was titled "What are Google Penalties? 7 Penalties & How to Fix Them" and worked through seven numbered types. Another page-one result was titled "20 Worst Types of Google Penalties You Need to Avoid" and worked through twenty. Both lists were counted that day. Neither set matches the report's entries, and the two do not match each other.

The mismatches worth naming, because each one sends a reader to the wrong fix:

• Doorway pages: both lists treat doorways as their own penalty. Google's list does not. Doorways appear only inside Thin content with little or no added value, first as one of the examples alongside thin affiliate pages and scraped content, then again in that entry's fix steps. They never get an entry of their own. Reading them as a separate action means looking for a report entry that will never appear.
• Hacked content: both lists include it. The Manual actions report does not, because hacked sites belong to a different report entirely. More on that below.
• "Pure spam": the second list uses this label for its ninth entry. The phrase does not appear on Google's page at all. The nearest entry is called Major spam problems, so a reader hunting for the label they read about will not find it in the report.

The deeper problem is a vocabulary one. The first result has a section headed "Algorithmic penalties", which treats a core update and a manual action as two flavors of the same thing. They are not, and merging them invites a specific waste: filing a reconsideration request over something no reviewer ever touched.

How do you check whether you have one?

Open Search Console and go to the Manual actions report. It is the only place a manual action is confirmed, and Google is explicit about what you will see: a count of manual actions at the top of the report, or a green check mark and a message if there are none. Google also notifies affected sites in the Search Console message center.

The Manual actions report is the first thing I open in an audit, before any crawl, because it is the one report that can make every other finding on the list irrelevant. If a reviewer has demoted the site by hand, the Core Web Vitals work can wait.

Where the report sits, and what the rest of Search Console is telling you at the same time, is covered in our walkthrough of the Search Console reports.

What a clean report rules out: a manual action. That is the whole list. It says nothing about an algorithmic reassessment, a deploy that shipped a stray noindex, a SERP layout change, lost links, or a seasonal dip. It is a genuine answer to one question, not a verdict on your traffic.

Does a manual action hit the whole site or only part of it?

Either, and the report tells you which. Google's instruction is to expand the manual action description to see a list of patterns of affected pages, and that this "can be a subset of your site, or the entire site." A pattern looks like https://example.com/real-estate/*. A site-level action carries the description "Affects all pages".

There is a caveat inside Google's own wording that is easy to skip past: "Not all of the pages matching the pattern are necessarily affected." The pattern tells you where to look. It is not a list of sentenced URLs.

Note also that Google does not use the terms "site-wide match" or "partial match" on that page. Those are industry shorthand for what the report expresses as a URL pattern or the phrase "Affects all pages". Useful shorthand, as long as you know the report will not be speaking it back to you.

Scope decides the size of the job, not the standard of it. A pattern covering one directory means you have a directory to clean rather than a site. The standard is identical either way, and Google states it flatly: fix the issue on all affected pages, because "fixing the issue on just some pages will not earn you a partial return to search results."

What does the Manual actions report not cover?

Hacked sites and malware. Those belong to the Security Issues report, and Google draws the line itself: the Manual actions report covers "manually detected issues with a page or site that are mostly attempts to manipulate our search index, but are not necessarily dangerous for users", while the Security Issues report covers "indications that your site was hacked, or behavior on your site that could potentially harm a visitor or their computer".

The practical difference is visible to your customers: security issues can put a warning label in the result or an interstitial in the browser, and manual actions carry no such signal. So "hacked site penalty" is a category error: the report you are watching will never show it. If the site was compromised, the report you need is the other one, and the work is remediation and a security review rather than a request about spam policies.

How do you fix the violation, by type?

Group the fifteen actions into four kinds of work, then follow Google's sequence for the one you have.

Link actions: Unnatural links to your site and Unnatural links from your site. Google documents an order for these, and the order is the part people get wrong: removal first, the disavow file only as the fallback for what would not come down, the review request last. Our guide to the test Google's disavow documentation actually sets works through that sequence and the harder question of whether to file at all.

Content and technique actions: thin content, major spam problems, hidden text and keyword stuffing, cloaking and sneaky redirects, cloaked images, sneaky mobile redirects, back button hijacking. The work is the same shape in each case: remove the pattern at the root across every affected page, not sample it. A template that cloaks one page cloaks every page built from it.

Platform actions: user-generated spam, site abused with third-party spam, spammy free host. The spam is someone else's and the action is yours, so the fix is moderation and prevention rather than deletion. A forum cleaned once but still accepting the same submissions fills up again.

Format and markup actions: structured data issue, AMP content mismatch. Bring the markup inside the guidelines, or bring the two versions of the page back into line.

Two rules apply to all four groups, and both come straight from Google. The first is the partial-fix rule above. The second is easy to overlook: a reviewer has to be able to see the fixed pages. Google says affected pages "should not require a login, be behind a paywall, or be blocked by robots.txt or a noindex directive", and points you at the URL Inspection tool to confirm it. A site that fixed everything and then left the staging noindex in place has fixed nothing a reviewer can verify.

What does a reconsideration request actually need?

Three things, and Google names them. A good request "explains the exact quality issue on your site", "describes the steps you've taken to fix the issue", and "documents the outcome of your efforts." The third is what makes the first two verifiable.

What each of those means in a request a reviewer can act on:

• The exact issue: name the action as the report names it and describe the mechanism that produced it. "We ran a link campaign through an agency between March and August, and those links match the pattern in Unnatural links to your site" is a statement a reviewer can check. "We think some old links may have been low quality" is not.
• The steps: dated and countable. How many URLs, which directories, what changed on each, what was removed and when. A list beats a paragraph, because a reviewer is checking coverage against the affected-pages pattern.
• The outcome: evidence, not assertion. Removal confirmations, the crawlable URL where the old content used to sit, the disavow file scoped to the links that stayed up, a before-and-after on a representative page. That is what separates a claim from something a reviewer can confirm.

Three more instructions sit in the documentation. If you have more than one manual action, Google asks you to read about and fix all of them, not just the one that scared you. If you bought the site after the violation happened, Google's instruction is to fix the issues and then say so in the request. And once the request is in, do not resubmit before you get a decision on the outstanding one.

You submit it from the report itself by selecting Request Review. Google's separate help page on reconsideration requests defines the term and points back at the report, which is worth knowing before you cite it for guidance it does not carry.

How long does the review take, and what if it is rejected?

Google's documentation for the report gives one sentence on timing, and it is a range rather than a deadline: "Most reconsideration reviews can take several days or weeks, although in some cases, such as link-related reconsideration requests, it may take longer than usual to review your request." You get an email confirming receipt and a second email when the review is complete.

That is the entire documented answer. A page quoting a specific window in days, or a number of months to a "full recovery", is not quoting Google, because the documentation supplies no figure to quote. Treat any precise duration as somebody's average rather than a commitment.

If the review goes your way, Google's wording is that it will revoke the manual action. If it does not, the report stays as it is. Neither the Manual actions report page nor the reconsideration requests page describes an appeal route, a rejection reason code, or a cap on how many reviews you can ask for.

That absence is useful. It means a rejected request is not a documented verdict on your case. Google publishes no reason code, so the only reading available is that the reviewer could not confirm the fix, and the loop runs again. Google's documentation sets three conditions a request has to satisfy before a reviewer can confirm anything, and each one is a way a request fails without anybody forming an opinion about your intentions:

• A partial fix: some affected pages cleaned, others missed. This earns nothing, by Google's own statement.
• Unreachable pages: the fixed pages are behind a login, a paywall, a robots.txt rule, or a noindex. The reviewer sees the old state or nothing.
• A second outstanding action: another entry in the report you did not read, did not fix, and did not address.

Check those three before you rewrite the request. They are cheaper to verify than the prose is to redraft.

What if the report says no issues detected?

Then you do not have a manual action, and the work in front of you is diagnosis rather than appeal. That is a real answer rather than a dead end: you have eliminated the one cause with a defined route back, which tells you the remaining causes all need to be identified rather than argued with.

The shape of the drop is the fastest way to narrow what is left. A sudden sitewide fall on the day of a deploy behaves nothing like a gradual slide across two weeks, and impressions holding while clicks fall points somewhere different again. Our guide to reading the shape of a traffic drop works through that sequence, and it is the right next read from here.

What not to do meanwhile: file a disavow, rewrite your best pages, or roll back a release as a precaution. A clean report means no reviewer acted against you, so a fix aimed at a reviewer has nothing to land on. When the guessing starts costing more than the drop did, running the diagnosis properly is what our SEO services are for.

The habit worth building is small: read the report's own words before you translate them into industry shorthand. The label, the affected-pages pattern and the phrase "Affects all pages" are the three facts everything downstream depends on, and every one of them gets lost in paraphrase. Before your next traffic postmortem, pull up the report and record exactly what it shows, with the date beside it, whether that is a named action or nothing at all. A dated clean check is what lets you say three weeks later that a reviewer was never involved, instead of trying to remember whether anyone actually looked.