Google gives different search results based on browser?
I was doing some searching on Google the other day and noticed that I got different search results when using Firefox and Konqueror. I thought it would be fun to do a little search experiment.
I downloaded a list of 138 user agent strings. Next I picked five words from Google's zeitgeitst (Bar Refaeli, brad delp, god of war 2, richard jeni and sinbad). I set up a perl script to run search queiries against google. The script would search each word 138 times, using each of the 138 browser user agents. Then I ran the script 10 times to make sure I got consistent results.
It looks like google gives two different search results. Which browser user agent used, help determine how many search results were returned.
Example. For the search "god of war 2", user agent " Mozilla/2.0 (Compatible; MSIE 3.02; AOL 4.0; Windows 95) " came back with 80,000,000 pages. For user agent "Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.0.6) Gecko/20060808" came back with 81,500,000.
For "brad delp", user agent "Mozilla/5.0 (Macintosh; U; PPC Mac OS X; it-it) AppleWebKit/103u (KHTML, like Gecko) Safari/100.1" gave 93,500 results, but user agent "Mozilla/5.0 (Macintosh; U; PPC Mac OS X; es) AppleWebKit/125.2 (KHTML, like Gecko) Safari/125.8 " came back with 1,340,000.
Running the query multiple times always gave the same results. I would have suspected that running the same search multiple times would have different results, since you probably bounce around googles server farm. But in this test, the number of results returned seemed to be tied to the browser's user agent. To make sure that nothing like cookies or browser security effected the results, I used perl and the LWP module to run the searches. The only difference between each search was the user agent that perl sent back to google.
Below is a sample of some of the results I got:
| User Agent | Search Word | Counts |
| Mozilla/2.02E (Win95; U) | brad delp | 1340000 |
| Mozilla/3.01Gold (Win95; I) | brad delp | 93500 |
| Mozilla/4.0 (Compatible; MSIE 4.01; AOL 3.0; Windows 95) | brad delp | 1340000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_68K) | brad delp | 1340000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_PPC) | brad delp | 1340000 |
| Mozilla/2.0 (Compatible; MSIE 3.0; AOL 4.0; Windows 3.1) | sinbad | 6250000 |
| Mozilla/2.02E (Win95; U) | sinbad | 6250000 |
| Mozilla/3.01Gold (Win95; I) | sinbad | 411000 |
| Mozilla/4.0 (Compatible; MSIE 4.01; AOL 3.0; Windows 95) | sinbad | 5920000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_PPC) | sinbad | 5920000 |
| Mozilla/2.0 (Compatible; MSIE 3.02; AOL 4.0; Windows 95) | Bar Refaeli | 458000 |
| Mozilla/3.01Gold (Win95; I) | Bar Refaeli | 458000 |
| Mozilla/4.0 (Compatible; MSIE 4.01; AOL 3.0; Windows 95) | Bar Refaeli | 470000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_PPC) | Bar Refaeli | 470000 |
| Mozilla/2.02E (Win95; U) | richard jeni | 938000 |
| Mozilla/3.01Gold (Win95; I) | richard jeni | 938000 |
| Mozilla/4.0 (Compatible; MSIE 4.01; AOL 3.0; Windows 95) | richard jeni | 918000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_PPC) | richard jeni | 918000 |
| Mozilla/2.0 (Compatible; MSIE 3.0; AOL 4.0; Windows 3.1) | god of war 2 | 80000000 |
| Mozilla/2.02E (Win95; U) | god of war 2 | 80000000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_68K) | god of war 2 | 81500000 |
| Mozilla/4.0 (compatible; MSIE 4.01; AOL 4.0; Mac_PPC) | god of war 2 | 81500000 |
Thanks for your Overview
I’ve been looking for a such tool for several months. Early I used intop20.com service, but it was only for USA Google results scanning. I’ve compared and positions were the same. Also I’ve asked a person who is in US now to look throw serp. And again - positions are the same as your plugin shows. (but they differ from positions in my country, surely). It seems to work good.
google
This phenomenon has always existed
I can confirm this happens
There's no doubt this discrepancy is a real effect. We regularly poll our ranking in Google and we consistently see different results in Mozilla than we do in Safari. We haven't regularly tested this with IE, but the Mozilla/Safari results are always different, they are also consistent per browser, so searches in Mozilla will regularly produce similar results, and a search in Safari will produce similar results, but a search in both will produce different results.
Google results definately different
I am not sure if it is a browser issue that gets different results posted but I was on the phone from Canada calling Connecticut and I asked my friend to check resuls for the key word "Koozies" on Google.com ...My site was ranked #1 in Conneticut on Google.com... but where I was sitting(Canada) on Google.com I was ranked #3 We both were using IE 6
That's simply a location
That's simply a location issue, nothing to do with the browser. I have several sites that do well in the UK but don't rank at all in the US, even though they're not locally focussed. This is mainly caused by the original and focus on the inbound links.
it's not just dumb luck. the
it's not just dumb luck. the author of this article is on to something.
i just installed Safari for Windows and it gets different results than i get in Firefox. that could be explained by some of the comments other people have made here. but get this - if i try the same search on my Powerbook, Safari/Mac shows the exact same results as Safari/Win, and Firefox/Mac shows the exact same results as Firefox/Win.
so there's no chance that it's related to cookies, caching, or preferences. it's definitely got to do with serving different results based on browser.
perhaps a secondary server is used to serve Safari results. Apple probably has a deal with Google, due to the built-in search box in Safari. maybe Google looks for the Safari user-agent and serves results from the server that's reserved for sending results to Apple users. this is made somewhat more complicated by the release of Safari on Windows, since the original deal was probably intended to serve results to Apple customers and not just Safari userrs.
do a search for "cvs commit"
do a search for "cvs commit" in IE and Firefox.. see for yourself.
Google testing
I know google does a lot of "live" testing - meaning that they'll take a certain reproducable section of the market, like 5% or whatever, and show them results from a different google algorithm (or a different design, etc). Then they can test and see how long you searched for that query, and whether one was significantly faster than the other.
Perhaps they were doing this when you were testing this - it' be interesting to see if the results held up their differences over time...
-Kevin
Re: Role of caching and the server farm
I thought about the caching and server farms. I repeated the same searches over the course of a few hours. I always got the same results. Meaning useragent A always got one set of counts and user agent B always got a different set.
That is still a possibility though, but that would mean the caching was based on user agent or that the server load balancing was based on user agent.
Role of caching and the server farm
How much consideration did you give to the possibility that:
a) your searches by search text & user agent would have gotten cached, so you would see the same results. To control for this, you could repeat the searches over several hours / days.
b) your queries went to different servers on the server farm complex. each one of them probably holds a slightly different version of the index as the index is always changing all the time.
Also, what if you add a level of control by initiating searches from different client PCs.
Post new comment