Link Velocity Spike

What a link velocity spike is supposed to be

A backlink is an inbound link from another website to yours. Link velocity is a practitioner term - not a Google one - for the rate at which a site acquires them: links per day, per week, per month. A velocity spike is a sudden sharp increase in that rate.

The claim this page examines is the negative SEO version. It says an attacker can harm a site simply by causing a large number of links to appear in a short time, because the rate itself trips an automated filter. In this telling the links need not be spammy, the anchors need not be poisoned, and the sources need not be bad. The speed is the weapon; unnatural growth is detected; the site is demoted for growing too fast.

The claim has to be stated precisely, because the precise version is the one that fails. Nobody disputes that a burst of identifiably spammy links is detected. The question is narrower and sharper: is volume-over-time, on its own, absent any spam signal, a documented input to anything at Google?

I looked in the three places such a signal would have to show up - Google's published documentation, Google's public statements, and the leaked internal API documentation - and did not find one. That is the finding, and everything below is the working.

The term does not appear in Google's documentation

Search Google's spam policies and Search Central documentation for "link velocity," "link growth rate," or any equivalent and nothing comes back. Google defines link spam by purpose and method - "the practice of creating links to or from a site primarily for the purpose of manipulating search rankings" - and never by rate. There is no published safe pace, no threshold, no advice anywhere to acquire links slowly. The absence is the finding.

Nor does rate appear where it would most naturally be raised. The disavow documentation is the page Google wrote for owners who believe hostile links are harming them, and it never mentions acquisition speed. What it says instead is that "Google works very hard to make sure that actions on third-party sites do not negatively affect a website," and that "In most cases, Google can assess which links to trust without additional guidance, so most sites will not need to use this tool."

The 2016 Penguin announcement is silent on rate while being explicit about mechanism: "Penguin now devalues spam by adjusting ranking based on spam signals, rather than affecting ranking of the whole site." The stated input is spam signals. Not speed. A folk belief this widespread leaving no trace at all in a decade of primary documentation is itself evidence about the belief.

The leak is the evidence people cite, and it says the opposite

Google's internal Content Warehouse API documentation, published to a public code host in March 2024, is the best available window into Google's own vocabulary for link handling. It does contain rate and spike fields, and those fields are what people point at to prove link velocity is real. Read the descriptions and they prove the reverse. From the anchor phrase spam module:

• phraseAnchorSpamStart - "Time when anchor spam spike started with padding."
• phraseAnchorSpamRate - "Average daily rate of spam anchor discovery."
• phraseAnchorSpamDays - "Over how many days 80% of these phrases were discovered."
• phraseAnchorSpamPenalty - "Combined penalty for anchor demotion."

The parent field in the anchor statistics module states the purpose outright: these signals "identify spike of spammy anchor phrases," and anchors created during the spike are tagged LINK_SPAM_PHRASE_SPIKE.

Every rate and spike field in the corpus is qualified by spam. What is measured is not the rate at which links arrive but the rate at which spam anchor phrases are discovered. The spike being identified is a spike of spammy phrases, not a spike of links. Spam classification is the precondition; timing is applied afterwards, to work out which anchors arrived inside the window so they can be tagged and discounted.

So timing here does the opposite of what the folklore says. It does not convict - it bounds. The successor module keeps the same shape and adds demotedStart and demotedEnd, documented as the start and end dates of a demotion period. The documented use of rate data is to fence off the period an attack occupied. That is a containment function, and from the target's point of view a defensive one.

Two honest limits. I did not enumerate the whole corpus; I examined the anchor and Penguin modules where such a signal would live, and found no field measuring raw inbound link acquisition rate independent of a spam classification. And documentation is not deployment: the phrase spam module is itself marked deprecated in favor of a successor, so even the spam-conditioned spike detection described here may be historic.

The documented cases rule rate out as the variable

Three public link-attack cases exist in any detail, all of them involving extreme rates, and they disagree about outcome in a way that eliminates rate as the explanation.

• 7 June 2012. TastyPlacement pointed roughly 52,000 links at its own thin exact-match-domain test site over days. The site fell out of the results.
• 30 July 2013. Nick Ker absorbed an escalating attack that reached 1,000 links a day against his established site, and reported "No 'unnatural links' warning in Webmaster Tools, rankings are nearly the same as they were back in February."
• 17 November 2014. Eliav Lankri described roughly 600,000 links, explicitly "spread over time," against an e-commerce client - and a penalty followed.

Line those up. Ker's case had a very high sustained rate and no effect. Lankri's was deliberately smoothed, low-velocity, and drew a penalty. The high-rate case survived and the slow one did not. Whatever distinguished the outcomes, it was not speed; it was the target's existing profile and the character of the links. These are weak sources individually - a vendor infographic, a victim's blog, a practitioner post - and they are used here for the pattern across them rather than for any single measurement. But if velocity were the mechanism, this is the one place it would have shown up, and it points the wrong way.

Why it never made engineering sense

This next part is an argument rather than a citation, and it is offered as one. If raw link velocity tripped a filter, every successful public relations campaign would be an act of self-harm.

A product launch, a story that reaches the front page of a news aggregator, an acquisition announcement, a citation from a major reference site, a viral piece of research - each produces a link acquisition curve far steeper than any plausible attack. None is known to cause a demotion, and it would be extraordinary if one did. Search engines exist to surface what people have suddenly started talking about. A system that penalized sudden link growth would be a system that penalized exactly the events it was built to detect.

That is why the velocity claim never made engineering sense, and it is worth holding onto when a graph looks alarming. The shape of a link curve does not distinguish an attack from a triumph. Any system reading rate alone would have to treat them identically, which is a sufficient reason for no such system to exist.

Where the folklore came from, and who keeps it alive

The claim is a fossil of a specific window. Between 2012 and 2016, Penguin ran as a periodic filter: rankings changed on refresh dates that Google controlled and did not announce in advance. Site owners watching positions move on those dates built rate-based folk explanations for outcomes that were actually driven by link quality and by a release schedule. The explanation was wrong but it was the only one available from outside.

Penguin 4.0, on 23 September 2016, moved Penguin into the core algorithm and made it real-time and granular. The December 2022 link spam update pushed detection further upstream, with Google stating that when "our systems nullify spammy links, the link credit that was previously generated is lost." The mechanism the velocity folklore was invented to explain has been replaced twice over.

The folklore outlived it because it is sold. Drip-feed link packages and velocity-smoothing services exist because the fear exists, not because the signal does. Some of the loudest published material asserting that velocity is a ranking input comes from firms selling the remedy, and some of the material denying it comes from link vendors with their own interest - the conflict runs in both directions and neither side is a source. What settles it is that nobody on either side can point to a primary document. Anyone selling you protection from link velocity is selling you protection from a signal no primary source describes.

When a spike really does precede a problem

Being precise here matters, because defenders genuinely do see spikes followed by drops. In every case I can identify a mechanism for, the mechanism is something other than rate:

• The spiked links are spammy. Then the spam is doing the work and the rate is incidental. That is a spam link attack, and it has its own page.
• The spike carries poisoned anchors. That is anchor text poisoning, and it is the one case where the leaked LINK_SPAM_PHRASE_SPIKE tagging genuinely applies - to the phrases, not to the volume.
• The spike is your own doing. A site that bought links and then blames the timing is describing a cause it paid for.
• The spike triggered human review. A conspicuous pattern can draw a reviewer's attention, and a reviewer who then finds real violations issues a manual action for the violations. The spike was the tripwire, not the offense.
• The site is brand new, with no history, where any signal is a large fraction of a small total.

None of those is a velocity penalty. Each is a different thing wearing a spike's clothing, and the triage question is always the same: what are the links, not how fast did they come.

What to do when your backlink graph spikes

1. Confirm the spike is real. A backlink index charts when a tool discovered links, not when they were created. Tools crawl in batches, and a large share of alarming spikes are artifacts of a crawl schedule. Compare against Search Console's own link data before anything else.
2. Look at the anchors, not the count. Search Console, Links, then Top linking text. If the wording is clean, the spike is a curiosity. If it is not, the relevant page is anchor text poisoning.
3. Check for a manual action - a penalty applied by a human reviewer at Google and shown to you in Search Console. "No issues detected" means no penalty was issued, whatever the graph looks like.
4. Check the update calendar against the date the traffic actually moved. In the large majority of cases the drop lands on a confirmed core or spam update and the spike is a coincidence you noticed only because you went looking after the drop.
5. Do nothing, and keep a dated record in case the pattern continues.

What actively hurts: velocity smoothing and drip-feed services; disavowing a spike wholesale; pausing legitimate outreach or publicity because a chart looks steep; and reconsideration requests filed with no manual action outstanding, which ask Google to reconsider nothing.

The damage this myth causes, and the recourse it does not have

This page exists to stop one specific act: filing a disavow file because a number went up. That act is the mechanism by which the velocity myth causes the damage it claims to prevent. A disavow file permanently discards link equity, and Google warns on its own disavow page that the tool "can potentially harm your site's performance in Google Search results" if used incorrectly, and that "most sites will not need to use this tool." Its stated criteria require a considerable number of spammy links and a manual action they have caused or likely will cause. A steep chart satisfies neither.

The second, quieter cost is worse: site owners who decline good publicity because they are afraid of unnatural growth. That mistake costs more than any attack described on this page, because it forfeits real gains to avoid an imaginary penalty.

As for recourse, there is effectively none specific to this vector, and that is the correct answer. Google has no process for reporting a link spike, and there is no harm to plead where no mechanism of harm is established. Every legal theory that reaches a link campaign reaches it through what the links say - a defamatory statement, a misused trademark, a false claim about a competitor - or through unauthorized access to a computer system. None of them reaches how fast the links arrived. Where a spike is one component of a wider sabotage campaign, the recourse belongs to that campaign, not to the graph.

Frequently asked questions