Welcome to WebmasterWorld Guest from

Message Too Old, No Replies

Synonym Handling in Google is Far Too Wide

You want Widgets? Here are results for NotWidgets, WideWidgets & Widgets!

8:17 pm on Oct 18, 2013 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Oct 9, 2004
votes: 0

I should state up front that I'd like any responses to this to be about the specific problem of synonyms being "wrong" - it's going to be tempting to veer from that track, hence the request to stay focused.

The Problem:

It seems that synonym (single words and phrases) identification is being experimented with to an extent which is causing relevancy issues (more than usual with synonyms).

Have you noticed a rise in the number of searches which highlight words that are not correct for the search you made?

I have tested the problem with thousands of phrases that are vertical specific (in 200 different verticals) and chosen ones which I have painstakingly hand verified for several indicators of specificity (to measure how "far" they are from each of the other phrases in the vertical).

My results show that Google appears to not understand that phrases within verticals can mean very different things and that these differences can even suggest a term is consumer as opposed to business targeted.

The problem seems to be at its worst when the phrase in question incorporates a word that describes the vertical and includes another word which is rarely found (1% to 5% of websites in vertical) which means 90%+ of websites in the vertical don't cover the subject/product/service. In this instance you often see wider phrases being matched against the search, and those "matches" are often very wrong - thus making the results poor.

To make matters worse, there are sites that have the specific phrase but Google is choosing a snippet which highlights the wrong phrase - so there's no hint which would allow someone to see that site is better in the snippet.


This is tricky without specific examples, here's what the phrases target.

"broad service (whole vertical) aimed at businesses"
"quite specific service (but popular in that sector) aimed at businesses"
"very specific service (rare in that sector) aimed at businesses"

"broad service (whole vertical) aimed at consumers"
"quite specific service (but popular in that sector) aimed at consumers"
"very specific service (rare in that sector) aimed at consumers"

Assume that the websites usually target either businesses or consumers but rarely both (10% target both).

The results of a fairly representative search show; on a search for "very specific service (rare in that sector) aimed at businesses":

1st and 3rd are "very specific service (rare in that sector) aimed at consumers" - which is of no use whatsoever, this has NO relation to the search.

6th, 9th and 10th are correct

2nd, 4th, 5th, 7th and 8th are
"broad service (whole vertical) aimed at businesses" which might offer the service because they do cover commercial services (but in reality none of the websites did).

That's pretty shocking, especially when I know that there are 8 websites that offer the service and cover the geography searched for - all 8 of those websites are in the Google index and none seem to be penalised (although none have strong link profiles, they are local business websites).

The Lesson to be Learned?

From an SEO perspective it's clear that you need to make sure the title of the page is going to show that specific service/products/subject as you cannot rely on the snippet (which is not in any way news, it's just getting even more important). This is only going to encourage more thin pages again if left as it is - Google needs to fix this issue, pronto.
9:42 pm on Oct 18, 2013 (gmt 0)

Senior Member from US 

WebmasterWorld Senior Member lucy24 is a WebmasterWorld Top Contributor of All Time Top Contributors Of The Month

joined:Apr 9, 2011
votes: 244

you need to make sure the title of the page is going to show that specific service/products/subject as you cannot rely on the snippet

That's assuming for the sake of discussion that the search engine* uses the page's own title (see assorted other threads going back a couple of years) instead of making up a new title more closely tailored to the search ... and/or to what it believes the search to be.

So far search engines don't seem to have messed with the meta description, so that's another place to concentrate. I don't know if anyone has collected hard evidence. But my nebulous impression is that if the meta description is a good match for the search, while the specific search terms occur on widely separated parts of the page, the search engine will go with the meta.

* I know the OP was specifically about google, but the same behavior has been reported in the #2 search engine.
10:11 pm on Oct 18, 2013 (gmt 0)

Senior Member

WebmasterWorld Senior Member 10+ Year Member

joined:Oct 9, 2004
votes: 0

Yes, you can't rely on anything really! As always, best to make sure anything that does get left untouched will do the trick.
9:12 am on Oct 19, 2013 (gmt 0)

Senior Member from GB 

WebmasterWorld Senior Member 5+ Year Member Top Contributors Of The Month

joined:Sept 16, 2009
votes: 48

I've seen this (albeit on a far more limited scope than thousands of keywords) in a vertical that I have a client in.

This is a B2B vertical and, for several long search terms that precisely describe B2B product set, Google fills most of the top 20 results with B2C, media and hobby results. Two results sets in particular are very bad, one to the extent that not one site in the top 20 has ever been relevant in the 6 months I've been tracking.

Here's my take away from this:

Google is choosing not to rely on title tag, nor headers, nor the keywords in a particular order in the body text. My clients (and one or two of their competitors) have sites that I would consider spot on relevant in 'old fashioned' Google, because the words are used, in an order that creates the B2B phrases, in their titles, headers, body text, links.

Google is missing this: why?

I believe that it's because the individual words used in the query are silo-ed separately, and because one or more of the words in these phrases also have very popular, far wider, B2C connotations. There are huge amounts of media, UGC and other 'fresh', 'popular' content around these keywords.

If Google thinks a page is a good general fit based on these metrics for several of the words in in the phrase (and perhaps the fourth word that completes the phrase is mentioned once in passing somewhere in the page, or in a link to it), then it plays safe / obfuscates traditional on page optimisation and returns the partial matches from the more popular silos.

I believe that suggested search has a good deal to do with this. The phrases are very low demand (but incredibly high yield, continent-wide or world wide infrastructure contracts result from them), so Google does not have enough clear user intent to go on.

The way round this? We're hoping it's more relevant authority. The site does place well for the main 'catch all' terms, but poorly for the long tail. We're doing very specific B2B media outreach with expert level content to flow back traffic / reputation to the specific pages (or at least the page group) rather than the home page, to try to convince Google that our B2B pages deserve a mention in their B2C results sets.
5:47 am on Oct 21, 2013 (gmt 0)

Preferred Member

10+ Year Member

joined:May 21, 2004
posts: 374
votes: 4

I have also noticed this. In particular I am noticing this with keywords which can have multiple meanings. In the past Google never had a problem with providing correct results, but now a search returns results that have absolutely nothing to do with the query. It is almost as if they are looking at each word individually rather than the query as a whole.