Why Google stopped telling you your links are spam

For years, the worst thing that could happen to your backlinks was also the most useful: a manual action. Google would tell you, in writing, inside Search Console, that it had found unnatural links pointing at your site. It was a punishment, but it was also a map. You knew what got caught, so you knew what to undo, and the sharper operators knew what to do differently next time. That feedback is mostly gone now, and I do not think that is an accident of pure benevolence.

The shift: from loud penalties to silent neutralization

The change has been underway for a decade. When Penguin became part of Google’s core algorithm in 2016, Google stopped demoting whole sites for spammy links and started simply devaluing the links themselves, in real time.[1]Google Search CentralPenguin is now part of our core algorithmGoogle’s 2016 announcement: Penguin became real-time and now devalues spam links rather than demoting the sites they point to.View source ↗ Then came the December 2022 link spam update, powered by SpamBrain, Google’s machine-learning spam system, which was built to detect manipulative links and neutralize their credit, again without a notice.[2]Google Search CentralLink spam update, now powered by SpamBrainGoogle: SpamBrain detects sites buying and passing link credit, then nullifies that credit. The links remain, the equity stops flowing, and no message is sent.View source ↗

Manual actions still exist. A site with a blatant, deliberate paid-link scheme can still get a written penalty, and Google has even folded spam reports back into that manual process. But for the everyday case, the bought guest post, the cheap niche edit, the link package, the default response is no longer a penalty you can read. It is a quiet shrug. The link sits there looking alive while passing nothing.

The provocative thesis: silence kills your feedback loop

Here is the part practitioners feel but rarely say out loud. A penalty is feedback, and feedback is the one thing an optimizer cannot operate without. Optimization is a loop: do a thing, measure the result, adjust, repeat. A manual action closes that loop cleanly. It says, in effect, “these links, this site, caught.” Reverse-engineering a filter is much easier when the filter keeps telling you when it fired.

Silent neutralization breaks the loop on purpose. When a link is quietly discounted, there is no signal. Rankings do not move, but rankings not moving has a hundred possible causes. You cannot tell whether the link was ignored, whether it is still pending, whether it helped a little, or whether something else entirely held you flat. With no error message, there is nothing to optimize against. You are debugging with the console turned off.

Read it from Google’s side and the incentive is obvious. Every manual action is a free lesson handed to the exact people trying to defeat the system. Replace it with silence and the lessons stop. The spam economy keeps spending on links that do nothing, link vendors keep selling inventory that quietly died, and Google never has to argue with anyone about a penalty. A filter that gives no feedback is a filter almost nobody can train against.

The steelman: it is probably convenience, not conspiracy

Now let me argue against my own spicy version, because the honest take is more defensible than the cynical one, and an opinion you can defend is the only kind worth publishing.

The boring explanation is mostly correct: algorithmic filtering simply scales in a way human reviewers never could. Google has said its machine-learning spam systems catch vastly more spam, far more efficiently, than rule-based detection ever did, keeping the large majority of search results spam-free.[3]Google · Webspam reportHow we kept Search safe in 2022Google credits SpamBrain and AI-based detection with keeping the overwhelming majority of search results free of spam, far beyond what manual review could cover.View source ↗ A team of humans writing penalty notices for every bought guest post on the web is never going to happen. Neutralizing links by algorithm is faster, cheaper, less appealable, and harder to game. Of course Google chose it.

So the feedback starvation is most likely a side effect, not a master plan. But, and this is the part I will defend, it is a side effect Google has every reason to be delighted with, and zero reason to fix. They are not going to add a “your link was just devalued” notification, because the absence of that notification is doing useful work for them. The intent does not have to be cynical for the outcome to be exactly what a cynic would design. You can hold both: the mechanism is scale, and the silence is a feature they are keeping.

“But I buy links and my rankings go up”

This is the strongest objection, and it deserves a straight answer instead of a dodge, because plenty of people have genuinely watched rankings climb after a link campaign. I am not going to tell them they did not see what they saw. But “rankings went up after I bought links” is not the same claim as “the bought links moved the rankings,” and two things usually sit in that gap.

First, SEO never happens in a vacuum. A team running a link campaign is almost always doing five other things at the same time: publishing content, tightening internal links, fixing technical issues, earning the occasional genuine mention, and simply aging and growing the site’s authority. When rankings rise, the bought links take the credit because they are the line item someone paid for, but the real attribution is tangled. The only way to actually isolate a link’s effect is a controlled test, links changed and nothing else, and almost nobody runs one. Correlation wearing a receipt is still correlation.

Second, the filter is strong but not perfect, and it was never binary. SpamBrain is probabilistic, not a wall. Some manipulative links slip through, especially the higher-quality placements that read as genuinely editorial, and those can still pass real value. So “bought links do nothing” is too strong a claim, and I am not making it. The accurate version is narrower: the expected value of a bought link has fallen, and, more importantly, become unknowable. You are not buying a guaranteed dud. You are buying a lottery ticket whose odds you are no longer allowed to see.

And that is the trap. A tactic that worked zero percent of the time would get abandoned fast. A tactic that works sometimes, for reasons you cannot see, funds an entire industry. Partial, inconsistent, unmeasurable wins are exactly what keep people spending. That is not evidence the bet is good. It is evidence the bet is sticky.

Why this makes paid links and disavow a worse bet

The practical fallout is the part that should change what you spend money on. If you cannot tell a working link from a quietly discounted one, then buying links becomes a bet you can never grade. You pay, rankings do not move, and you have no way to know whether the link was neutralized or simply never mattered. The vendor has every incentive to tell you it is “still indexing.” You are funding inventory that may have died on arrival, with the receipts deliberately withheld.

The same silence drains the disavow tool of most of its purpose. Google’s John Mueller has been blunt that, without a manual action or a history of deliberate manipulation, maintaining a disavow file is largely a waste of time, because the algorithm already ignores the links you would be disavowing.[4]Search Engine JournalJohn Mueller on the disavow toolMueller: for sites with no manual action and no deliberate link manipulation, disavow work is effectively a billable waste of time. The algorithm already discounts those links.View source ↗ Gary Illyes has gone further, saying he would remove the disavow tool entirely because it hurts more sites than it helps, since owners routinely disavow perfectly good links on the say-so of third-party “toxicity” scores. Bing already retired its version. The whole toxic-backlink and disavow industry is, in large part, selling a cure for a disease the algorithm already treats for free.

What to do instead: earn what cannot be silently switched off

If the game is now rigged against feedback, stop playing the game that depends on it. Google can flip a switch and nullify a link’s credit in silence. It is far harder, and far more conspicuous, to nullify a genuine editorial citation or a branded mention sitting inside real coverage. So move your effort to the things that do not have a quiet off switch.

• Earn editorial links and mentions, do not buy link equity. A journalist citing your original data, a writer naming you as the source, a real resource page choosing to include you. These are expensive to fake and awkward to silently discount.
• Build brand mentions, not just backlinks. Mentions correlate with AI-answer visibility far more strongly than backlinks do, and they keep working even where a link would be devalued. That is the connected play I cover in the AI visibility guide.[5]Ahrefs · 75K brandsWhat correlates with AI Overview visibilityAcross roughly 75,000 brands, branded web mentions correlate with AI-answer visibility about three times more strongly than backlinks.View source ↗
• Measure outcomes you can actually read. Rankings, traffic, and citations for the pages you care about. Stop grading individual links, because Google made that grade unreadable on purpose.

The uncomfortable takeaway is that Google removed the scoreboard for link manipulation, and a lot of the industry is still playing as if it can see the score. The operators who win from here are the ones who stopped trying to beat a filter they cannot get feedback from, and went back to earning the kind of authority that does not come with an off switch.