Google Results Platform Dependant?

22 comments

After observing this for a number of weeks and unless I'm completely in left field, Google results are platform Dependant.

[edit - added]

Quote:
So here I am going to present two screenshots search term used is “sports”, one from my Mac Powerbookng Firefox.0.4 and one from my Windows based Dell Lattitudetop using Firefox.0.4 - as you can see for a number of terms Googlep roducing very different SERPS based on the operating system of the client machine.

I am not technically competent enough to judge this one, perhaps someone with the knowledge can chip in. Cornwall [/edit]

Comments

links busted

1) your link to the second serp is broken
2) neither link is evident in firefox. On my system your links look exactly like the rest of the text. (there are two image links in the text, but it's like trying to find waldo).

1) The author needs to

1) The author needs to change his link color to something noticable

2) Trim down those log files so you can actually see what you're trying to show

3) I could not recreate this at all. For the most part, I got the exact same results on my Mac vs. my PC. When it was different, it was the normal datacentre fluctuation that we've been seeing for weeks (months). There are at least 4 different result sets showing for "sports" across the DCs. I was never served anything specifically based on the machine I chose.

The first screenshot, link corrected: http://www.flickr.com/photos/dancouver/168511008/in/set-72157594167865777
And the mac ss: http://www.flickr.com/photos/dancouver/168505908/in/set-72157594167865777/

I concur

One of my sites, xMule.ws is a fairly mature cross-platform opensource filesharing utility; it's very atypical traffic from most of what you guys probably get. Btw, I invest absolutely zero time or effort in SEOing this site, except for the "Bush resigned" query :-)

  • 60% non-U.S. (40% Europe, 18% Asia, 2% Other)
  • 81,687 search referrals for June, 2006
    • Google (86.13%): 72,158
    • About.com (7.16%): 6,118
    • Yahoo (3.45%): 1,754
    • Others (3.26%) - 1,657
  • Operating Systems
    • Unix deriv: 53.01%
    • Windows: 41.30%
    • MacOS: 5.48%
  • Browsers
    • Mozilla: 65.31%
    • IE: 19.56%
    • Konq + Safari: 6.87%
    • Opera: 6.32%

The amazing thing is that numerous people had reported finding different links based upon browsers, particularly around ED2K-specific queries. See, the vast majority of ED2K users are on Win32 (via eMule); a significant minority (maybe 5% probably less) are non-Win32 of which my program captures maybe 50% of the total searches.

Cross-checking Operating System to search keyword shows that it's roughly 2x more likely for a visitor to be in Firefox and a non-Win32 OS to come via a query with the word "emule" in it. For the highly-competitive keyword "p2p" (responsible for 2% of all referals) there seems to be a 10:1 non-Win32:Win32 ratio.

So, I would say that in specific, platform-dependent searches for highly competitive keywords (+10 million), there does seem to be a measurably large biasness, but one which I can't just plop down an example query.

You can sometimes pull up

You can sometimes pull up different SERPs with IE than with Firefox on the same computer.

Must be Google

going for SERP cloaking again ... :-)

Google Contradictions

Google cloaks regionally all the time and will also penalize other sites for cloaking. So I agree, I think this is one of those instances.

What do you think the result would be if I cloaked based on OS and Google found out? Would I get delisted? Would I be penalized?

Quite likely.

No they're not!

The Windows search:
http://www.google.com/search?hl=en&q=sports&btnG=Google+Search

For the Mac:
http://www.google.com/search?hl=en&q=sportsbook&btnG=Google+Search

Notice that they are different keywords used in each search...

That was my mistake ...

in the headers logs I posted. The searches for the same two terms on mac and pc are consistently giving different results and at times IE versus Safari produce different results.

This is platform cloaking without a doubt and I'd like to get a decent response on why this would be the case. I see very little reason why platforms shold produce differing results - in what way is this more relevant for the user?

I can post further examples across numerous terms where this is the case.

Sorry, but I still don't see

Sorry, but I still don't see it happening in my test. My only suggestion is that Google picked your computers for some of their random testing? Who knows. Which keywords are you searching for btw? Would love to try out a few!

Has anyone else seen this?

Post is updated

I've updated the actual post Google SERPs Post in addition to the Mac Header Log and the Windows Header Log and separated those logs from the post.

I even highlighted the proper link info as well...hey it's K2 and has a few bugs ok?

Cheers,
Dan Nedelko

many do

but cloaking based on OS is excusable on the grounds of usability, in most cases.

I'm sorry but I can't see that this is the case, not only because I can't reproduce it (are you sure you're not hitting different datacentres? Have you tried using an IP for the search rather than www.google.com? Are you logged into your Google Account on one machine?) but also because there's absolutely no point - it would be true that different OS users' are likely to have slightly different demographics which may make some results marginally more preferred on one OS than on another, and in theory they could be checking that out, but thats soooo marginal given that the majority of office workers have no choice in their OS, and anyway they'd be stupid to be doing that in the middle of all the (self acknowledged) issues that they currently have because any satisfaction data thet get would be meaningless at this moment.

And while I think they can behave in a less than phd-intelligent way, I can't believe they're that stupid :)

My views summarized

No one apparently read my above post... People searching for a cross-platform filesharing *network* are many times more likely to land on my site (a non-Win32 p2p program for said network) if they are using a non-Win32 computer. Roughly 20:1 for a 100 million-results query (p2p) which I have done 0 SEO toward.

I read it

but don't you think that it might be because more people using a non-win32 computer are interested in a cross platform file sharing network on a site called xmule? Or would search for p2p as opposed to "way to share files across platforms on a network"?

Mostly cross platform networks are administered by the people using the non-win32 computers rather than the ones who are I guess :)

Not too surprising, really

Though we haven't analyzed this phenomenon to date at greater depth, I don't find it altogether surprising when held against what's generally known and verifiable re Google's overall corporate and philosophical makeup.

To wit: Google's patently not primarily a search engine, as it has led so many to believe (including probably 90% of all media hacks, even the technically savvier ones who should know better), but first and primarily a data mining outfit, very probably even the world's largest, if we discount a few of the better known state run intelligence agencies. (No coincidence, of course: Google's head honchos' NSA connections are well established, even though the pertinent documents proving it were hastily pulled from public access when this factoid finally hit the grapevine a couple of years or so ago.)

Data mining being a generally very little understood concept and technology, it's perhaps understandable that falling for their comparatively harmless "we're just a greedy search engine making mega bucks by inundating the web with AdWords/AdSense and hypocritically complaining about click fraud and people churning out content spam to jump on our bandwaggon, even though we made most of this possible in the first place" spiel is so predominant everywhere.

But let's not forget that you'll hardly find any other outfit on the planet with more detailed and gapless, uninterrupted highly specific (and not even particularly granular) data on individual Internet users' (think: Google toolbar, GMail, WebAccelerator, Writely, and as of recently Spreadsheet) online behavior going way back to the mid nineties (at least as far as Yahoo! is concerned) than the search engines. The same applies to surfers' mass bevavior patterns, of course.

So who's to say if they haven't found out that the kind of browser or platform specific traffic you're attracting isn't exactly what's statistically typical for and pertinent to your kind of site? Or that they aren't conducting some experiments on these lines? Look at what an excellent job AdSense is generally (yes, there ARE exceptions, but not too many) doing now of determining what any given webpage is actually about to serve contextually relevant ads? And how fast they're doing it.

I'm aware that this doesn't address your question in as detailed and explanatory a manner as you might prefer, but IMV anything else would amount to mere divination.

However, if any of this should prove to be true, it may hold a lot of surprises in stock (not all of them too pleasant, I'm afraid) for us SEOs. Admittedly, much of it is mere surmise at the present point, but is it really implausible if we assume that all that hullabaloo currently encompassing linkage, PR, sandboxing etc. etc. will soon subside to give way to an entirely different search ranking paradigm: namely, one based on surfers themselves, starting with their established search behavior, their hardware, their browsers, their surfing history, their e-mail communication (all of which is being monitored automatically, remember, under the pretexts of serving ads and catching terrorists in one fell swoop - a very American melange, I if may say so ...), their individual text output and stored personal files, their financial calculations, their media viewing preferences, etc. etc. - in short, their whole frigging micro demoscopics?

What this would mean in practical SEO terms is that - always assuming that they'll actually succeed to a greater extent than incurring failure in this venture - the current, "classical" SEO methods of remote search results manipulation are doomed for the simple reason that the customary model of "on site" factors (the fondly nurtured "content is king" myth, plus page structure and layout, etc.) and "external" elements (such as incoming and outgoing links, IP location, domain registration data, etc.) will be considerably demoted if not entirely superceded by what are essentially the instruments and mechanisms and processes of social engineering, mass behavior analysis and crowd control.

Which would only leave us with the ghastly realization that good old Orwell can, in retrospect, only be faulted for having been such an inveterate optimist ...

Fantomaster ...

..Amen. A well written argument. I particularly like your conclusion

Quote:
Which would only leave us with the ghastly realization that good old Orwell can, in retrospect, only be faulted for having been such an inveterate optimist ...

nice post fanto

Quote:
What this would mean in practical SEO terms is that - always assuming that they'll actually succeed to a greater extent than incurring failure in this venture - the current, "classical" SEO methods of remote search results manipulation are doomed for the simple reason that the customary model of "on site" factors (the fondly nurtured "content is king" myth, plus page structure and layout, etc.) and "external" elements (such as incoming and outgoing links, IP location, domain registration data, etc.) will be considerably demoted if not entirely superceded by what are essentially the instruments and mechanisms and processes of social engineering, mass behavior analysis and crowd control.

Which to me says you hire the SEO *individual* and not the "firm". You hire a creative and intelligent business person, with the skills and knowledge (resources) to get the job done. You hire talent, or muscle, or genius, within context.

"remote search results manipulation" is replaced by "remote search engine manipulation" and "remote user perspective manipulation".

Egads, I may someday actually be a "search engine optimizer" after all.

Search behavior perception manipulation

might be one of the options, but of course it would require truly vast resources to effectively skew the engines' statistical data in any meaningful (and exploitable) manner.

E.g. create a global network of clickbots simulating "search personalities" (possibly millions of them: browser specs + cookies + JavaScript etc. implementation included) and daisywheel them across an even vaster network of IPs (not sure if the current inventory of anonymous proxies would suffice - apart from the issue that they're fairly easy to pinpoint) to create specific artificial "user group search patterns" which you and/or your clients' sites are optimized for.

Don't forget blog comments (did I hear someone shout "spam"?), forums, social bookmarking, Gmail, Yahoo! Mail, Alexa, etc., good old Usenet etc. and whatever else may evolve in the near future.

In short, dock onto some real life demoscopic target group as closely as possible (provided you have their core data to work from in the first place - not a trivial issue!) and gently, gently feed your manipulative input into that specific data set to dilute it in your favor. Then, if successful, that target group will be fed your sites by preference as being the "most relevant".

Conceivable? Maybe. But actually doable? Possibly, however, the micro-variables involved being almost unlimited, this approach merits a big, fat question mark.

it is going to be more viral

it is going to be more viral / buzz marketing than just getting citations

even the linguistical patterns will likely become highly important... ie: what words are your brand commonly used near and related to? who uses them and where?

Absolutely!

IMV, the current performance quality of AdSense proves that they've finally come to terms with Latent Semantic Indexing (LSI) and similar linguistic approaches big time.

Conceivably, the immediate next step will be to scale this up to handle billions of documents, taking into account their intrinsic "social networking" patterns (aka linkage) until the point when this factor can finally be dumped in favor of demographics driven flagging as that seems a lot harder to manipulate from the outside.

So until that latter tech finally kicks in, SEO will be all about corpus linguistics, topic or theme specific semantic fields, etc. etc. Merely synomizing your stuff to escape duplicate content detection won't suffice anymore: sure it will evade that particular issue, but when tied into a social search pattern framework it will only be one of your lesser, more trivial concerns as an SEO.

Instead, you'll have to make 100% sure that your pages' overall semantic content is in total accord with whatever the search engines algo deem to be appropriate to your topics. (Of course, this doesn't address the issue of multi-themed pages - which, interestingly enough, is the one area where AdSense still fails miserably.)

Seems they have to increasingly resort to human editing/sanitizing of SERPs - which, in a way, is a highly regressive approach throwing us back into the early 90s when Yahoo! popped up with its directory. However, we can be dead sure that they're not at all happy with that (far too slow, too human labor intensive, too suceptible to editorial errors, corruption issues, etc. not to mention the costs) - so they'll do all they're capable of to minimize their dependence on it.

Obviously, there's no small irony in the fact that you can actually exploit Google's AdSense as a prime optimization tool - seen from this angle, it's even niftier than Infoseek used to be back in the nineties. Talk about beating them with their own weapons ... :-)

(Of course, this doesn't

(Of course, this doesn't address the issue of multi-themed pages - which, interestingly enough, is the one area where AdSense still fails miserably.)

I am a bit mixed on that. I have seen some multi-theme pages do just pathetically with AdSense, while seeing others that appeared quite impressive (when they had at least somewhat similar core themes).

The biggest thing to note with AdSense (as you mentioned) is that it is pretty advanced in general, and far far ahead of the offerings from the likes of Yahoo!.

I'll keep it shorter than

I'll keep it shorter than fantomaster. :) We don't score documents differently based on a user's reported OS.

DanN, I'd clear your cookies on each computer and try running those searches again.

I see someone technically competent did chip in

[edit]..I am not technically competent enough to judge this one, perhaps someone with the knowledge can chip in. Cornwall [/edit]

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.