google has become very complex I would say.
If I understood the theories discussed here correctly then it might not just be anchor text or text surrounding the anchor text, but all the on page factors of the pages linking to a site. Something like: If the majority of the pages listing the keywords of the query agree that this page is an important source then it must be number one.
And: if the majority of the pages linking to this page come from countryx and the user querying for it comes from countryz then it is not that important.
Within that there is a lot of room for doing things differently and my personal explanation for a lot of the flux is that google is testing various possibilities and their effects.
For what I am watching the title of a page seems to carry a lot of weight.
Well, Google may consider text surrounding links, but the anchor text is very important as well. Ambigous link texts make accesibility problems (see WCAG 1.0) and indicate poor information architecture.
The actual text in the <a></a> tag is very important. It is hard to judge about the surrounding text as it is harder to measure. A simple test of the <a></a> tag is easy to create, but the subtle effects of surrounding text make a definitive answer elusive.
Has anyone actually run a detailed comparison, and wish to share?
Associating backlink anchor text with the linked site/page is much much easier than doing so with "surrounding text".
Already you can see the problems of determining what's relevent to the linked site in this context.
If you're talking internal site/on-page optimization, the role of anchor text vs. non-anchor text may be worth considering, but as far as external backlinks, I'd say its too much of crapshoot.
Google is counting on people to link intelligently, something they don't always do, but to extract and analyze the surrounding text of an external backlink is just asking too much. That's getting deep into theming/semantics. Better I think to just get the anchor text and look at the on-page factors of the linked site to determine relevence to particular queries.
actually I shouldn't think it to be too difficult if the Search Engine sticks to the keywords of the query - there is no semantics involved and I agree - google does not do any semantics yet, you even get different results with hyphenated or non hypenated keywords.
The nearness of the keywords to each other and to the links might count, and a lot of other factors including page rank.
|actually I shouldn't think it to be too difficult if the Search Engine sticks to the keywords of the query |
How does that help? You still have index every word of whatever you define as "surrounding text" and associate it with the linked page. The keywords of the query are a real-time input. Google can't guess what you're going to search for next week and it can't re-analyze this kind of stuff on-the-fly for every query.
I've tested anchor text v surrounding text.
Anchor text won.
I question whether Google considers surrounding text at all with links.
Did you notice any affect of surrounding text?
Yes but no where near as much as the anchor text.
peewhy, could you explain how you tested? Which variables you isolated? ;)
Sure, I've "tested" lots of stuff, but done nothing in a qualitative, analytical manner that would hold up to scrutiny. Since you threw yourself out there as an "algorithm tester" please tell us your procedure! It would enlightening for us all.
>rfgdxm1: I question the whole issue of link text.
And I question you, how many PageRank papers have you read? Ones authored by Brin & Page & others, such as Kleinberg, Herzinger, Bharat, Golub (do believe that's how to spell his last name) - which papers led you to question the experience of countless webmasters?
Or your tons of time posting here most especially in the Google News forum I've read tons of your posts - why are you questioning what 99.9% of webmasters take as fact of Google life?
Do enlighten us what led you to question these truisms, posts without basis most easily confuse those with limited attention spans such as myself ;)
Jeremy_Goodrich is up for confrontation again;)
|Sure, I've "tested" lots of stuff, but done nothing in a qualitative, analytical manner that would hold up to scrutiny. Since you threw yourself out there as an "algorithm tester" please tell us your procedure! It would enlightening for us all. |
Neither did I throw myself as an "algorithm tester" nor am I taking your bait.
You know how to test links and compare which works against others - single another victim out of the arena.. this one isn't playing :)
Well, you said you knew which one worked better. Instead of throwing out random ideas, perhaps you could share with us a bit of your knowledge? ;)
You said previously, "I've tested anchor text versus surrounding text. Anchor text won."
Again, how did you test? What # of pages? PageRank of the links? Number of results returned for your target phrase? Just curious, mind you, or are you simply quoting from the famous anatomy of a large scale hypertextual search engine paper, which is the 1st time in 1996 that they (Brin & Page) mention that "anchor text" is a key factor in their ranking algorithm?
After all, if you aren't testing ranking algorithms, then what are you doing comparing the results of rankings based on one variable versus another in the Google search engine results pages?
You are being confrontational. I'm sure others out there will be happy to play with you.
You have an excellent way of getting people not to share knowledge
I'm being analytical & objective, friend. Trying to "raise the bar"...and, besides that, anybody can go & read some Google research published at citeseer, stanford, etc & gain some pretty intimate knowledge of how Google works.
When we throw out ideas in a public forum, it's very important not to make generalized, non specific statements that can lead others astray ;)
After all, if the information you found here was "low quality"...then, would you have stuck around? Of course not!
Btw, Golub is the last name of a researcher at stanford who's doing some amazing things with adaptive pagerank, calculations on the fly, etc - but that's just numbers.
The anchor text & contextualization of various linking methodologies can only be determined by accurate heuristics, and I was simply trying to ascertain whether you had those or not in hand. :)
Well, an easy test for me was to take a piece of random text that is already next to a link on my page. Search for it and set &filter=0 in the google url.
No pages showed up from my site except those the exactly had the text.
SOooo... does this prove anything? Dunno. Maybe it will only give a boost if the text exists on the target page - can't prove that though.... need absolute balanced PR,etc for that.
[edited by: jcoronella at 7:29 pm (utc) on July 22, 2003]
|I always thought that anchor text was way overrated. |
I think the Exact opposite ;)
I believe the original paper by Larry and Brin talked about Text surrounding the link. And It can be Anchor text So anchor Text is the Text surrounding the link IMHO
I initially thought that <title> attribute of anchor Element might be the Text surrounding the links (it can be closest to the actual link) but then remembered Tedster saying in a thread something to the effect of "done lots of experiments with a variety of unique terms used in text close to the links (ie in <title> attribute ) and still those texts did not Show up for any results" (not his Exact words mind you, only the way my tiny Brain understood it to be).
If you feel that anchor text is overrated compare the results of allinanchor: with normal google results and you should see they are nearly identical in Competetive categories. But then Past few months quite a few things have changed and my assumptions might not be true now.
And I feel more than surrounding Text the Title of the Page (from where you get backlinks) holds more importance. Would love to hear what others thnk :)
mil2k, I'll have to read your post a couple of times because I'm not sure what you meant. But to clarify what I meant in case I wasn't clear, I thought the "concept" of using only <a href=>THE ANCHOR TEXT</A> and NOT the surrounding text is overrated, I wasn't referring to what Google is actually doing because I'm not really sure.
|I'm not sure what you meant. |
Yes may times I don't communicate things properly. Sorry My mistake. Also I misunderstood your question.
|I thought the "concept" of using only <a href=>THE ANCHOR TEXT</A> and NOT the surrounding text is overrated |
You mean ppl should go for organic links? If that is the case then I tend to agree that many ppl are missing out on organic links.
It's way too late out here. Will follow the thread in the morning:)
I'm not the Great communicator either :) I'll try again. I was talking about what Google SHOULD do vs. what they actually do. I think a search engine would be better off analyzing surrounding text as much as the link text itself.
OK, time to get some sleep myself, everything will be clear after some snooze :)
|I think a search engine would be better off analyzing surrounding text as much as the link text itself. |
How would directory listings, eg, interact with such a contextual algo? Would you inherit some of the tail end of the description that precedes you alphabetically in ODP, or would they give more weight to the text that follows a link (as when the directory description follows its title)?
I've done some anecdotal observation only... no testing... on a very rare misspelling that was adjacent to an inbound link, and I can say that I saw no boost whatsoever to result #100. Again, there were so many interfering factors that it really was a non-test.
To measure an effect that is as small as I'm assuming this one might be, I think you'd have to find a target term that's very rare to start with so it's completely non-competitive... perhaps make one up... and then use surrounding words that are also rare or made up. It would be interesting to see whether there's any effect at all, but my guess is that if it's very hard to measure, it's not going to be worth the effort. Again, this is complete conjecture.
>To measure an effect that is as small as I'm assuming this one might be, I think you'd have to find a target term that's very rare to start with so it's completely non-competitive... perhaps make one up... and then use surrounding words that are also rare or made up. It would be interesting to see whether there's any effect at all, but my guess is that if it's very hard to measure, it's not going to be worth the effort. Again, this is complete conjecture.
That I have never seen a single real world example of this is why I doubt surrounding text has any relevance. I'd have expected if surrounding text was being considered, someone would have been able to cite an example that proved this. This would likely involve where the surrounding words were rare, or made up. That I haven't seen a single example that made me suspect this leads me to conclude Google doesn't consider surrounding text.
Yo, dude! If you're gonna say you did testing, you'd better be prepared to follow up and describe the kind of testing that was done. Someone saying they did testing implies that there were "control factors" and a testing environment, and people accustomed to testing algos can generally classify and quantify those, so when they read such statements they expect what we could call scientific inquiry.
I don't need or want science; I want results that replicate themselves based on the same methodologies. I do not consider SEO an exact science, I consider it a creative art - which is why I love it.
If I see something work I try it someplace else and again someplace else, and if I get the same results consistently, that's quantitative enough for me. But I don't claim to have done testing. I am the Accidental_SEO and I like it just fine that way. It's been working quite nicely for a few years, and if it ain't broke I don't fix it. But I'm personally generally careful to qualify statements in order that what hasn't been quantitatively analyzed doesn't get inordinately construed as intended to be authoritative, which is never the intention on my part, or the part of many others.
There are, however, some who make what appear to be authoritative statements that are without substantiation. In order to maintain the integrity of the system and the reliability of the information presented, it continues to be incumbent upon those who make such statements to divulge the theoretical bases of their theorem.
Back to the topic of this discussion:
So -- surrounding text in close proximity to anchor text has absolutely_no_bearing on the ranking of the linking out page or the linked_to page? We're stating that as an absolute fact?
How about all the other elements on any given page that aren't text that's in immediate promimity or within the anchor text itself? They're all worthless?
Sorry, I really do need convincing, and always fully appreciate whatever enlightenment I can glean. I simply cannot buy all that. But then, what do I know? I'm not the WFA, I'm just the Accidental SEO.
Let's try to do some reverse engineering and analyze some keywords and backlinks if we're given permission (or we can do this over on the supporters forum). Anyone have suggestions for keywords?
Dunno about surrounding text but we tested a month ago <a href=>kw1kw2.com</A> with 3 links and our new 1 page site shot up to top 5 serps for search with kw1 kw2. Must be said that Google gives much weight to url's containing kw's. For search with kw1 only site is #25 and #300 for very competitive kw2.
Clark, there's a limit to what we can do out here in public forums, and there always will be. It just isn't the same thing.
mundonet, it works for sure up to a certain point, but if there's a great amount of repetiion - does that actually help, or does it start to go inversely related as far as the proportion goes?
What "great amount of repetiion"? Sorry I don't understand your question. There's no surrounding text in our test.
I dunnno if the good results come from the anchored text or because Google gives too much weight to domains with kw's in them but for sure the combination works. Because most of the time people linking to you are lazy and simply use the domain to link my conclusion is get a domain with the kw's in them.
I dunnno if the good results come more from the anchored text than because Google gives too much weight to domains with kw's in them but for sure the combination works. Because most of the time people linking to you are lazy and simply use the domain for the link, my conclusion is get a domain with the kw's in them. We are thinking of switching from our branded domain to kw1kw2.com because of how easy it was for a content thief competitor to get #2 right below us with kw1kw2.info. Maybe off topic, sorry!
-tired-of-seeing-advert-signatures-when-name-of-poster-is on-the-left-panel-next-to-the-post (you inveterate optimizers :)
[edited by: mundonet at 6:50 am (utc) on July 23, 2003]
|mundonet, it works for sure up to a certain point, but if there's a great amount of repetiion - does that actually help, or does it start to go inversely related as far as the proportion goes? |
I know Marcia what your concerns are. I have personally given it a lot of thought. Some of my possible thoughts were :-
1) Whether the links are reciprocal or one way. [at present both are working fine according to my experience. I have done some link work for a few Big companies and for one I got 1 way links for other I got reciprocals. Both are #1 for allinanchor:keyword and both are in top 5 for competetive and Very competetive terms.]
2) How do I prepare myself if google suddenly decides to tackle Identical anchor texts. Should I change some attributes of my link like Font Face, Size etc to make it slightly different than other incoming links?
3) Dilute the Keyword a Bit by adding a few words?
4) Avoid a standard HTML text put up on my Links page which people use to link to me?
These are just my personal thoughts and experiences :)
Sadly Google does not offer all the advanced searching Alltheweb does.
Fooling around with the ATW it seems ATW does nothing with surrounding text.
Searching for "the" whilst having the criteria "must not include" the in all the options - in the text, in the title, in the url, in the host name, in the link to url.
Of course ATW could decide to only count surrounding text if it exists in the target page...
| This 65 message thread spans 3 pages: 65 (  2 3 ) > > |