Hacker News Reader

Comment by demetris | Hacker News Reader

Matt, I don’t understand why Google need to offer blocking as a user option. Any web netizen with some experience can tell at a glance the spammy and low-quality results on a SERP, but Google, with their vast knowledge and experience and their ability to monitor all sorts and aspects of user clicking behaviour, cannot?

View on HN

I alluded to that in http://news.ycombinator.com/item?id=2218627 . People feel comfortable with Google removing blatant spam: hidden text, cloaking, sneaky JavaScript redirects, etc. People tend to feel less comfortable if they feel like Google is making an editorial decision.

If we get a good signal from this extension, or from offering block links in Google's search results, then it's much more similar Gmail's spam algorithm, where an email is labelled as spam partially because a lot of users say it is, rather than because of some editorial decision on our part.

Matt, before we get to the point where potentially controversial editorial decisions will have to be made, I would imagine there are things that could be done automatically and uncontroversially.

For example, sometimes we see copies rank higher than originals. Why does that happen? Google know where they first saw a particular piece of content, don’t they? Why don’t they use that as a heavy ranking factor?

Or am I too far off?

Consider this: I make a blog post on my relatively new blog examplesite.com. Techcrunch picks up on the article and immediately reposts it on their site. Which do you recon would be the first to get indexed?

Now which is the original from Google's point of view? Relatively smaller blogs can take significantly longer to index then sites that have massive amounts of content moving about daily.

I have thought about that and I cannot believe it is really a problem.

My small negligible personal website notifies Google, Bing and Yahoo immediately and automatically as soon as I publish something new. It also publishes a feed. Even if the content is picked up and republished right away by a site that is indexed every minute, it should be possible to determine correctly the original publisher.

In some cases, I can think of more ways to determine the original publisher. And certainly Google can think of even more.

> I can think of more ways to determine the original publisher.

Please share them. Thanks.

Here is one scenario that comes easily to mind:

I publish an article at only-original-content.com. The article has some images that are served from only-original-content.com/images. Now only-copied-content.com takes my original article and republishes it. Since only-copied-content simply copied the HTML, the images are still served from only-original-content.com/images.

In that case it should be simple to determine who is the original publisher. Of course, only-original-content.com could simply be a CDN that only-copied-content.com uses for its static resources, but, again, it should be easy to determine whether that is the case.

You might want to take a look at this:

http://news.ycombinator.com/item?id=2152286

demetris, that's exactly what we improved with a recent algorithm change: http://www.mattcutts.com/blog/algorithm-change-launched/ . As you point out, that was a more straightforward change, and that's why we were able to launch that one first.

That said, if you wanted to share some examples where you're seeing copies rank higher than originals I'm happy to pass that on to the right folks. In fact, some of the right folks are already on this thread. :)

posts to google groups show up in devcomments.com, osdir.com, mail-archive.com, etc ranked higher than the original post in google groups.

Hmm. That could be if Google Groups has changed the url structure, which could make crawling it harder. Or because USENET/mailing lists don't always have a centralized/canonical location on the web, which makes dupe content more of a potential issue. It's not the usual "Website X copied my website" scenario.

Getting on a tangent here, but Google has a hard time crawling Google Groups? Have you tried emailing support@google.com? Just kidding but in all seriousness how is it that a Google property has bad SEO?

still the sites mentioned above are arguably of lower quality than google groups, there is no reason to rank higher...

I'm really glad this extension exists.

However, if it is built in to the general results, could you also add a metric to Webmaster Tools showing the frequency with which your domain is reported. It would also be good if the blocking could have timeout period so that sites can be given a chance to improve their behavior, rather than just deleting that domain forever.

+1 to a report in WMT - to detect problems in our sites and also potentially abuse.

I understand the fine line you have to walk, and I'm glad to see that you guys take it as seriously as you do.

Personally, I'm glad that you're putting this out as a user-controlled thing. I like the fact I'm able to get rid of results that aren't necessarily spam or SEO'ed garbage, but I where I still know that I never want to see results from that site again.

You already have the SafeSearch filter that can be toggled on and off to show you different search results. Why not an Editorial filter as well (perhaps disabled by default)?

A "web netizen" is intelligent. Google - as you're using the term - is a sophisticated piece of software that leverages much of what we know about AI, but it is not actually intelligent. The determination you're making that something is spam and not worth your time may be simple for you to make, but it's hard to get software to consistently make the same determination.

It's one thing for an individual user to ask not to be bothered by results from an entire domain. It's quite another for Google to make that call for EVERYONE. I think this is a great first step. If the evidence is overwhelmingly against a domain, I would hope that Google would use that as a strong negative signal against the domain to adjust the ranking of the site.

Another thing to keep in mind is to think about how many unique search queries are entered into Google on a daily basis. I don't know what the figure is at Google, but some engineer at Bing said that 25% of searches they've never seen before. You can't get a human to look at each of these searches and remove the content farms, you have to do it algorithmically or else it will never work.

Getting users to do the work will help with the most egregious abuses. One thing that content farms are good at is coming up with answers for questions that don't have an answer available on-line. For example, the query "what's the personal cell phone number of <insert celebrity>?" will just take you to a spam farm because that's the only kind of website that claims to know the answer.

Reply on news.ycombinator.com