I actually have no idea what Google Analytics looks at to decide the user agent.  Certainly not my server logs ;) It might throw away the declared User-Agent entirely and rely on its own flavor of browser-sniffing, on the theory, why bother with the User-Agent at all if you're going to run your own tests for all the other attributes:
https://developers.google.com/analytics/devguides/reporting/core/dimsmets/system


-----Original Message-----
From: Code for Libraries [mailto:[log in to unmask]] On Behalf Of Robert Berry
Sent: Friday, July 13, 2012 7:37 AM
To: [log in to unmask]
Subject: Re: [CODE4LIB] Browser Wars

To be fair, I think this information isn't likely to be *that*
inaccurate. Most people don't count 'the User-Agent header of your HTTP
requests' alongside their age or income.

MJ Ray <[log in to unmask]> writes:

> Genny Engel wrote:
>> Interesting, Safari has just pulled into the lead over here.
>
> You can't know that.  With so many browser and proxies mangling the
> User-Agent for various reasons (User-Agent Switcher to get a nicer
> mobile-style experience on a small screen, or randomUserAgent to stop
> the evil empire tracking you through browser fingerprinting, to give
> two examples), reading the User-Agent header from your logfiles is a
> suggestion or hint of what's reading your site, not a definitive
> result.
>
> What's in the logs is basically reader-submitted.  You don't believe
> people to all tell the truth when they tell you their age or income,
> so please don't believe them about their browsers!
>
> I predict we will see much more volatility in these results as
> more people install the obvious plugins to get a nicer and safer
> browsing experience.
>
> Regards,