Really struggling to find tools that report latency loading a website from multiple geo-graphic locations.
Are there such tools? I have begun replicating infrastructure in other EC2 regions, but find it difficult to ascertain what performance improvements the regions users are seeing.
Update
What you are looking for is Real user monitoring (I've indeed misread your question initially) - it is certainly possible to implement your own solutions via a combination of the various Open Source monitoring frameworks available, but I'm not aware of a readily available one; however, at least two (likely several more) commercial providers are offering just this and have free tiers available one way or another:
New Relic is offering active Real User Monitoring, which is an excellent tool to Get browser performance data directly from real end users and especially correlate these with other monitoring metrics for the ultimate insight in your operational and development optimization efforts:
Cedexis is offering Cedexis Radar to capture the experience of your actual visitors:
Integration
Most interestingly, both offerings can meanwhile be integrated into Cedexis' other product Openmix, see Cedexis Adds New Relic Support To Its Real-Time Cloud Routing Platform for details.
Disclaimer
As easily figured from my phrasing, I've used New Relic with great success for the use case at hand (and lots of others) and can certainly recommend it without hesitation, whereas I've only started to look into the Cedexis platform and am not able to judge their offering yet accordingly.
Initial Answer
I have recently provided an answer to How could I determine which AWS location is best for serving customers from a particular region?, outlining two approaches easily available for Amazon EC2 specifically:
Please see the answer itself for more details on these two options (including pointers to background and additional information) - if you can't solve your problem right away via one of these approaches, the implied information might offer a good starting point at least.
You could use various looking glasses around the world, however automating this would probably break their ToS.
For automation, I really couldn't say without having some equipment/vm's in the areas or some sort of client side scripting to report latency.
If you're willing to spend the money, there's a company called Gomez.. well I guess it's Compuware now.
Compuware Gomez
They offer synthetic (backbone testing from dozens of locations worldwide) testing as well as 'last mile' and 'Real user' testing (where you have a .js beacon that monitors a % of your real traffic and reports back).
The monitoring can be a simple http page request or include all the subcomponents and displays a firebug-like waterfall load time breakdown.