31

The way Stack Exchange cross-links sites via "Hot Network Questions" is causing Google search results to be pretty useless... so I'd like to request that "Hot Network Questions" be removed from all sites or some other change be executed to help Google post more relative results.

Case in point:

The pages that are not part of the "ExpressionEngine" Stack Exchange site show up in my search results because, I assume, they link to the ExpressionEngine Stack Exchange site in their "Hot Network Questions" section in the right column.

Screenshot of the section:

My Google search string was the following if you want to test the results on your end:

"Maximum function nesting level" "expressionengine" 
16
  • 2
    If you're trying to find results for the ExpressionEngine site, you should use the advanced search operator site:expressionengine.stackexchange.com, like this.
    – Laurel
    Dec 5, 2017 at 0:42
  • 2
    Why do you assume HNQ has anything to do with it, rather than the fact that the results actually contain the string you searched for?
    – Cai
    Dec 5, 2017 at 0:55
  • 1
    @Laurel Though this is an option, this isn't realistic with the amount that I search. I also don't just want Google search results from SE... but I want valid search results from SE when they do show. Dec 14, 2017 at 17:57
  • 3
    @Cai This isn't an assumption. These CMSs are not related to each other and one wouldn't find "expressionengine" related content on the Joomla or Drupal sites. The crosslinking is polluting the search results. I've used Google as my main search method for years now and the results coming from SE have become less and less relevant when searching exact terms specific to the CMSs I work with. Dec 14, 2017 at 18:04
  • 3
    @Anna it is an assumption, you don’t know Google’s algorithm. My point is that all of those results literally contain your search string. You’re searching for a generic PHP error that shows up in all of those CMSs. The search results you show look perfectly normal to me.
    – Cai
    Dec 14, 2017 at 18:09
  • 4
    @cai I am searching also for ""expressionengine" which is not a term that shows on those pages except via the cross-linking. I can't think of any other way Google would be connecting the two exact search terms together with those CMS page results... can you? Dec 14, 2017 at 19:52
  • 1
    I really don't know, but links between to those sites doesn't seem like an obvious assumption. Remember those sites are all on the same domain so Google, AFAIK, will treat those as internal links and give them nowhere near as much weight as regular backlinks.
    – Cai
    Dec 14, 2017 at 20:27
  • 1
    also worth noting that if you replace "expressionengine" in your search with "wordpress" or "drupal" you get more accurate results, so I'd assume there are simply less ranking results that relate to ExpressionEngine
    – Cai
    Dec 14, 2017 at 20:29
  • 1
    @cai I would agree that internal vs external links have different weights... but internal links still have weight and if there is an internal cross-link, why wouldn't that generate incorrect results? Dec 15, 2017 at 0:06
  • 1
    I'm reaching this conclusion because my Google search is looking for exact matches of both terms... they are wrapped in quotes. So technically both terms "should" be present on the page, right? I'd argue that when Google crawled the site, the two terms were on the same page because of the crosslinking. Dec 15, 2017 at 0:11
  • 1
    While you have a valid issue here, the real solution for the example you've given is to use the advanced search operator site:expressionengine.stackexchange.com, as @Laurel said. It is certainly an option for you, particularly if you are doing a large amount of searches. It's easy to create a bookmark to Google's Advanced Search which automatically fills in the site you're interested in, example for expressionengine.stackexchange.com. You then just click on that bookmark when you want to search that site.
    – Makyen
    Apr 7, 2020 at 21:17
  • 4
    Seems more like the real solution would be to tag the Hot Network Questions block on the page for any given post, to exclude it from being index by Google, making the indexed keywords for that page more specific to the main content. This way we could avoid the index being "polluted" with keywords from whatever transient HNQ content happened to be present at the time of indexing. Allowing Google to index HNQ on a question page seems to be very net-negative: I can't think what the positive impact(s) of it might be? Jan 17 at 22:30
  • 2
    @TimWilliams I couldn't agree more. This continues to be an issue with Google results. I'm honestly surprised this wasn't fixed when I reported it 5 years ago. Jan 18 at 17:45
  • @Anna_MediaGirl - I can't believe I only just noticed it... Jan 18 at 17:47
  • @TimWilliams In all fairness, the search results on Google do seem better. I just tested a couple searches. So maybe it was fixed in StackExchange? Or maybe Google fixed things on their end? With which search term are you seeing this? Jan 18 at 21:07

2 Answers 2

15

I can prove this is an issue:

Hot Network Question titles are indexed and then show up in google searches, causing SE to pollute the results... in this example, 8/9 results on the first page are various SE sites and only one of them has anything to do with the quoted text; the rest only include the title in HNQ. This is particularly clear when you see the surrounding text in the quote, which is just other HNQ titles.

Google search that is just a page of random stack exchange websites: sci-fi (the one that relates to the term) as well as german, crypto and stackoverflow

Seems like it would be prudent to remove HNQ for search spiders.

Also worth mentioning is the only non-SE result is cloning content from the HNQ on SE.

5

Just ran into this myself.

Was trying to find an answer of mine on Stack Overflow so used the following search in Google...

site:stackoverflow.com "phil" react context

FYI, I'm using Google because Stack Overflow's search is next-to-useless.

This returned many posts matching my required terms but they weren't mine. In fact, the matched term isn't about programming at all and visiting the pages no longer contained the text.

Google search results

Each post refers to...

Does Phil. 2:10-11 describe Jesus Christ as God?

I can only assume then that Google has indexed the Hot Network Questions which at the time had a link to this post on Biblical Hermeneutics.

6
  • 1
    While I agree that SE's search leaves a lot to be desired, the specific case of limiting a search to a single user's posts and/or only answers or only questions is one where SE's search does perform reasonably well, sometimes better than general search engines, particularly where the username you're looking for is both short and common. In this case, the SO search for user:283366 is:a react context produces 7 results.
    – Makyen
    Feb 27 at 5:42
  • 1
    @Makyen I can never find what I'm looking for that way. In any case, this was mainly a demonstration of why HNQ is a problem for external search engines
    – Phil
    Feb 27 at 5:44
  • Re "the matched term isn't about programming at all": It isn't very clear. Perhaps add some annotation to the picture, like some free-hand red circles? Feb 27 at 6:03
  • @This_is_NOT_a_forum I didn't think it was necessary because a) it's emphasised in the results via bold text, and b) I noted the actual text under the image.
    – Phil
    Feb 27 at 6:06
  • 1
    I initially reported this 5 years ago and the fix is fairly simple. The Hot Questions area can be loaded dynamically using an ajax call AFTER the page finishes loading and this will hide the content from Google's search bot. I just submitted a support request asking the SE to look at this issue. Feb 27 at 18:38
  • 2
    Support suggested I post to the meta without apparently reading my message which include the link to this post... head slap Feb 28 at 19:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .