Google May Trust Links Less After They Are Modified

Jul 2, 2014 • 8:39 am | comments (32) by twitter Google+ | Filed Under Google Search Engine Optimization

Google LinksPedro Dias, a former Google employee who worked on the Search Quality and Webspam team for years, said on Twitter that Google is less likely to trust a link after it is modified.

He wrote:

Did you know Google is less likely to trust a link once it has changed from the 1st time it was seen?

I tried to probe Pedro on how he knows this. Was it from a test he conducted or was it from his past work experience at Google? My impression, but I do not have him saying this on record, was from his past experience:

So how much does Google drop the trust in the link? David Naylor asked Pedro and he implied, not too much:

He even added that less trust, in this case, is not a bad thing. I don't see how that is possible:

Anyway, this is a former Googler saying this who worked closely in web search. He did not say if this is based on his past work experience with Google or via new tests he is running based on him now being an SEO consultant.

Either way, I am sure the SEO community will have a field day with this.

Forum discussion at Twitter.

Previous story: Google Structured Data Testing Tool Also Drops Authorship Image


Steve Erlich

07/02/2014 01:20 pm

That's incredibly interesting Barry, and makes complete sense based on statements from Matt Cutts. If you are editing a link, you probably know it's there, you know it's coming and you know what it says. 3 Red Flags that this was a link built by SEO and less likely a link that occurred naturally. Kind of irritating though considering how many times webmasters have messed up links, or had a typo in a brand, or pulled data from other locations that then had to be updated. I mean there are many reasons a link can change, not just to best suit SEO.

David Eaves

07/02/2014 01:33 pm

I thought as much, links are like wine, they get better with age. It doesn't surprise me that changing a link could mess that up a bit.

Dave Fogel

07/02/2014 01:42 pm

Bad Links are Like Bad Wine... They turn to vinegar with age.


07/02/2014 01:52 pm

Its probably as simple as once it has changed, Google doesnt intrinsically trust the link like it did before, so it has to be there a while going to a trusted website to regain the trust it once had. It makes sense for there to be a trust threshold, where if the combined trust of the linking and linked to sites exceeded that, they don't get any more trust bonus, its just considered 'trustworthy'. If they did keep the same trust, it would be a nice loop-hole for hackers to use, by switching links on sites to their spam and keeping the inherited trust.

Josh Zehtabchi

07/02/2014 02:30 pm

All jokes aside, this makes perfect sense. A website has a set of NATURAL links, SEO is hired, SEO reviews links and says "Hey, they're using your company name, let's change that to your keyword" and a natural link becomes a synthetically modified link with the only purpose to rank higher.


07/02/2014 02:37 pm

I call BS... the data and analytic requirements to store and calculate reputation for every link would be way too high... its hard enough to do it for all sites and even harder for all pages, but for present and past versions of links as well... whats the upside for google besides flushing money?

Joshua Price

07/02/2014 02:41 pm

seems like it would be best to hire the SEO while you're in the drafting process of your client's site.

Josh Zehtabchi

07/02/2014 02:42 pm

Well, I don't even mean that. Let's assume the company that hired the SEO had some great natural links from various trusted sources and the SEO comes in to 're-optimize' said links. I've actually have seen proposals from SEO's who will do this. *shrugs*

Joshua Price

07/02/2014 02:47 pm

For me, I mix age with relevant keywords based on the customer's search. I do SEO for industrial tools, and my technique alone has not negatively impacted us; our scientific site has gotten traffic now; usually get one web order a year, now we're getting 4-5 per month. It's all about simple unique content, relevant keywords, a good sitemap and every now and then; direct url submission lol. hasn't failed me yet!

Steve Erlich

07/02/2014 02:48 pm

Very clever. Haha. Love a little corny SEO humor in the morning.

Steve Erlich

07/02/2014 02:51 pm

Oh absolutely. I won't name brands, but I followed a pretty large SEO agency who was a competitor at the time and watched them successfully maintain that link rotation. They simply figured out how often each page was indexed and setup their rotation accordingly. Could very easily do the same thing with malicious links or spam.

Luke Linkwalker

07/02/2014 03:08 pm

Those aren't seo's that go out and forensically change good inbound links to exact match, those are hacks. Good seo's create and repair. 2 sides to the force.

Don Dikaio

07/02/2014 03:10 pm

Barry do you know if this only applies to outbound links? How would this affect internal linking?

Josh Zehtabchi

07/02/2014 03:14 pm

Fair statement, let me leave it as self-proclaimed SEO's ;-)

green mile

07/02/2014 04:38 pm

I've seen a ton indian blog networks that rotate old links for new customers and they work great

Michael Martinez

07/02/2014 04:50 pm

This sounds like a hint to the "repurposed link" crowd who love to search for dead links.

Michael Martinez

07/02/2014 07:45 pm

The upside for Google is that they can see how the normal Web behaves and use that as a benchmark for monitoring suspicious link activity.

just sayin

07/02/2014 09:56 pm

I have always suspected things like this. Believe me if you were in Googles shoes and you wanted to identify the group trying to spam / scam you, and you have all the data and resources they have ... would you not include the most basic common sense ones? Repeated tweeking a site to try and gain higher SERP certainly could be identified, changing the back links is definitely one indicator and it is one of the ones most web seo guru's advise to do.

Jaimie Sirovich

07/02/2014 11:56 pm

They know when a link is no-followed. Why would they not know if it's no-constant?

Jaimie Sirovich

07/02/2014 11:58 pm

They can just update a 1-byte counter on each link to see how many times the link has been "tweaked." 0 is false. > 0 is true. 255 changes and it's worth zero. This is cheap to track. They don't have to store the anchor text — just how many times it changed in the same DOM position — though that could be tricky to track. Any link that keeps changing is subject to a formula that deprecates it with some sort of decay function. Or it just resets it to a "new" link but without a freshness boost if one supposes links appreciate value with age.

Jaimie Sirovich

07/03/2014 12:08 am

Pretty sure fresh links also get a temporary boost. So what Google might do when anchor text != anchor text is just reset the age of the link but not give it a boost for freshness. They're just going to do whatever's cheap and models reality without opening it up for gaming.

Jaimie Sirovich

07/03/2014 12:15 am

They could appreciate or depreciate. A timestamp and an increasing function would easily model this. Or a counter for how many times the anchor text changed in a decreasing function. I'm not shocked. This makes sense. Pretty sure that loop-hole would peter out as Google computes the slower stuff.

Gregory Smith

07/03/2014 05:27 am

Broken Link Building is just 1 of the modern day link building strategies that's used per editing links. I was thinking the same think Michael.


07/03/2014 06:10 am

Now I am more sure they are not doing this.

Jaimie Sirovich

07/03/2014 06:13 am

Why is that? The data needed are small.

Faisal Jamal

07/03/2014 12:23 pm

I don't think this is the case .... how Google knows that the link that is replaced more important then previous one .... sometimes links which are replaced more important than previous one... However, might be possible in case of plenty of modifications needs more explanations on it

Harish Bali

07/03/2014 02:26 pm

yes - i agree clear hint for people who keep an eye on expired domains to get benefit of authority built it over the years.


07/03/2014 02:58 pm

There are so many edge cases you might as well track trust with random numbers... what if a link moves on page. What about when there are multiple links to the same place. What if one is deleted or moves. What if it reappears later but is otherwise unchanged. What if 2 URLs are different strings but canonicalize to the same place according to the URI specification. What if 2 different links go to the same page because one is a redirect. What if 2 different URL produce the same content because the app renders it that way. Not to mention all the links that just naturally change on an active white hat site. This idea's chief output would be collateral damage.

Eric Ward

07/03/2014 04:29 pm

Here's your answer: It depends and here's why: What changed? The anchor text only, while the undelying URL remained the same? The URL, but not the anchor text? Both? And what type of change was it? A change of 250 anchors from "click here" to "payday loans" is quite different than a change of 7 links from "2014 International Hearing Loss Conference" to "2014 International Hearing Loss Conference and Expo". And how credible was the linking site in the first place? A change to an outlink on CNN's homepage versus a change to an outlink on Earl's Online Casino" site might be interpreted differently. Google has hundreds of PhDs. They are way smarter than me. If I can think of these scenarios, so can they. And they also have 15 year's worth of backlink statistical data they can use to create models of what a normal vs. abnormal link change rate should look like. And dead links get replaced all the time, some with prodding, others without, and some (likely) with a payment. There can be no possible one-fits-all scenario where a change to a link automatically means less trust.


07/03/2014 04:56 pm

Exactly right.

Nick Ker

07/03/2014 05:37 pm

I had to chuckle at the phrase "normal web", but good point.


07/09/2014 03:41 pm

There are a lot of factors at play, namely the quality of the site you are receiving a link from. If a link is over-optimized within a site's overall link profile with commercial anchor text, then I'd consider the link itself more on the low quality side of the equation and I'd likely change it. What would you have to lose? If you don't change it, then you might face Penguin. Which is worse? Losing a little trust or being mauled by Penguin?

blog comments powered by Disqus