Bot footprint: Only HTTP request or browser emulation?

Bots are very easy to spot, i think they just don't get baed as often as you'd expect because the websites dont want to risk baing any legit users. for example

Http parsing- easily spotted, what browser doesn't load CSS and images from the server, or execute JS
PhantomJS - loads everything and runs everthing but I can found from JS enumeration and analysis of headers (not user agent)
IE/Chrome/FireFox - Now if you can simulate real users pattes this is be best, however that can be very hard!

other things to consider..rendering pages you'll need a lot of resources for loads of requests, so it might not be economically feasible. I found this with an old twitter bot I wrote with PhantomJS. I would just go with HTML parsing and try not to set too many spam flags off....