Google manual actions and penalties, told apart

Traffic fell off a cliff and the word that comes to mind is “penalty”. Before you act on that, it is worth knowing that Google has two entirely separate mechanisms for suppressing a site, that only one of them is called a penalty, and that most sudden traffic drops are neither.

Check which one you have — this takes two minutes

Open Search Console and look at two reports. Security & Manual actions → Manual actions is the definitive answer on manual actions. It either names an issue and the affected scope, or it says “No issues detected”. There is no ambiguity and no hidden third state.

Also check Security issues in the same section, because a hacked site produces warnings in search results and traffic collapse that look like a penalty and are not one.

If both are clean, your drop is algorithmic, seasonal, a tracking problem, a technical fault, or the result of a competitor improving. Which is worth establishing before you start rewriting content, because the fix for “Google stopped indexing us after a migration” is nothing like the fix for a quality demotion.

Manual actions: what actually triggers them

Manual actions are issued by human reviewers against Google’s spam policies. The ones ordinary sites hit, rather than the ones aimed at deliberate spam operations:

Unnatural links to your site. Paid links, link exchanges, or a directory-and-forum-comment campaign, whether you ran it or an agency did. This is by far the most common manual action for legitimate businesses, and often the result of SEO work commissioned years earlier.

Unnatural links from your site. Selling links that pass PageRank without rel="sponsored" or rel="nofollow". Guest posts with keyword-anchored commercial links count.

Thin content with little or no added value. Scraped content, doorway pages, mass-produced pages that differ only by a city name, or auto-generated text. The bar is whether a page exists to serve a reader or to exist.

User-generated spam. Comment sections and forums overrun by spam you have not moderated. The action is against your site even though you did not write the content.

Structured data problem. Markup describing things that are not on the page — most often review or rating markup a business applied to itself. This one does not suppress rankings; it removes rich result eligibility, which is a narrower but very specific loss. See what is schema markup for the rule it violates.

Cloaking and sneaky redirects. Showing search engines something different from users. Occasionally accidental, when aggressive bot-detection or geo-redirection catches Googlebot.

Fixing a manual action

The process is genuinely a process, and the reconsideration request is the part people rush and should not.

Fix the cause completely, first. Reviewers check. A partial cleanup with a promise to finish gets rejected, and each rejection costs you another round trip of weeks.

For link actions, remove before you disavow. Genuinely attempt removal — contact the sites, document what you asked and what happened. Then disavow what you could not remove. A disavow file submitted without any removal attempt reads as refusing to do the work, and the disavow tool is not a substitute for cleanup.

Then submit a reconsideration request that says three things: what was wrong, what you did about it specifically, and what prevents recurrence. Include evidence — the spreadsheet of contacted domains, the dates, the outcomes. Do not argue that the action was unfair, and do not submit a second request while one is pending.

Review takes days to weeks depending on the action type; link actions take longest because they require the most verification. If the request is rejected you get a reason, and the honest reading of a rejection is almost always that the cleanup was incomplete rather than that the reviewer erred.

Algorithmic demotions: no letter arrives

Everything else works differently. There is no report, no notification and no reconsideration request, because no human decided anything about your site specifically. Systems that assess quality run continuously, and your site is being evaluated the same way every other site is.

The practical consequences are worth stating plainly. You cannot confirm you were affected — you can only observe a drop and reason about timing. You cannot appeal. And recovery is not a switch: improvements are picked up when the relevant systems next reassess, which can take months, and recovery is never guaranteed even when the changes are genuine.

The useful move is comparative, not introspective. Look at what now ranks where you did, on the queries that lost the most. If the winners are substantively better documents — more thorough, more clearly authored, actually first-hand — that is the gap to close. If they are not, check whether the query’s intent shifted, or whether you lost something technical.

Rule out the boring explanations

Before diagnosing a quality problem, eliminate the causes that are far more common and much easier to fix.

You de-indexed yourself. A stray noindex, a robots.txt rule shipped from staging, or canonical tags all pointing at the homepage after a migration. Search Console’s Page indexing report shows this immediately — see Search Console page indexing errors.

Analytics broke. A tag removed during a redesign produces a traffic graph that looks exactly like a catastrophe. Cross-check Search Console impressions, which are independent of your tracking code.

It is seasonal, or it was one page. Compare year-on-year rather than month-on-month, and check whether the loss is site-wide or concentrated in a handful of URLs that lost a featured snippet.

A migration lost redirects. Old URLs 404ing instead of redirecting discards their accumulated signals. Check for a spike in “Not found” in the Page indexing report.

What to do this week

Check the Manual actions report. If it is clean, check Page indexing for a technical cause, then compare your losing pages against what replaced them.

Our free site check covers the technical half of that list directly — whether your pages are indexable, whether canonicals are coherent, whether robots.txt is blocking anything it should not. It cannot tell you whether you have a manual action, because only Search Console can, and it is worth being skeptical of any tool that claims otherwise.