Google Search Classifier Went Haywire and Messed Up The Results

Crazy Robot Went HaywireOne of the worst things that could happened to a site owner is a drop in Google’s search rankings. As the most powerful driving traffic source on the web, every little change on its algorithm or how it is evaluating specific sites have a great effect.

Recently, one of Google’s signals (classifier) simply went bananas which apparently caused a huge change in the search results and affected many websites all over the world.

It all began a few days ago on April 16th. All worked more or less fine until suddenly at some point in early morning, all hell broke loose- Websites disappeared completely from the search index, other websites suffered from a huge decline in Google’s incoming traffic while many crappy sites flooded the search results.

SEO and marketing forums quickly flushed with an almost unprecedented amount of threads (I only recall this kind of amount after the first few Panda updates last year), reporting about all sorts of Google rankings related issues and crying out for some support to help identify the problem.

Take for example this forum thread from the Warrior Forum. Some person has reported that 19 out of his 20 sites have completely de-indexed from Google! Soon enough, many other site owners have posted that they are also experiencing with similar problems.

Here’s another example of a thread from Digital Point Forum- A person working for some SEO firm has reported that many of their clients are suffering from a major drop in rankings for their main keywords. Even though many tried to assist him, no one could really identify what caused it.

At first, many put the blame on the usual suspect, the Panda. Although the last Panda update occurred three weeks before and it seems a bit early for another update (mostly happens every 4-6 weeks), it appeared like a reasonable bet.

Another common assumption was that the overly SEO algorithm that was announced by Matt Cutts on March and expected to make a significant effect, may have finally arrived. The problem with all those assumptions was that the results clearly became worse as many spam and low-quality pages received more prominent presence on the search results.

What the fu** was going on?

As it finally reveals, it wasn’t Panda and neither any new algorithm. It was caused by a relatively small search classifier that renegaded… Matt Cutts disclosed it on a Google+ post:

I saw a recent post where several sites were asking about their search rankings. The short explanation is that it turns out that our classifier for parked domains was reading from a couple files which mistakenly were empty. As a result, we classified some sites as parked when they weren’t.

I apologize for this; it looks like the issue is fixed now, and we’ll look into how to prevent this from happening again.

Parked domains, if you don’t know, are essentially empty websites which in most cases only presents ads. Obviously when Google identifies a website as a parked domain it will be consider as a low quality site and probably won’t receive any search traffic at all. Therefore, this is the reason many legitimate sites have been wiped out (de-indexed).

It also had a significant collateral damage on other sites- The sites that Google identified as parked domains and de-indexed have linked into other sites as well. Meaning that a lot of legitimate link-juice other sites received has been deleted as well.

I’m glad that Google caught this problem only in couple of days, even though I’m sure that many site owner would probably feel the couple of days are a long time. Right now, I’m beginning to see a shift back to “normal” but it appears that the restoration has yet to be over.

Note: If your site suffered from a drop in rankings BEFORE April 16th, then it isn’t related to the current topic.

Advertisement