Here is a reference in English to the Colorfoto article:
The Colorfoto tests are in German. Unless you can read German you're going to have a hard time understanding the methods used in the tests. Rol_Lei Nut reported with specificity their conclusions. Take them for what they are-results of tests. If you can't examine the tests themselves that doesn't make the test results useless. I take them as interesting information, and my not being able to read German does make me dependent on someone else telling me what criteria the conclusions were based on. Without knowing the magazine I am also dependent on someone else telling me if in general their tests are well done. Still, in a test of this sort, results should be straightforward, as they measure things which can be well quantified e.g., "focused/unfocused" and fairly well quantified e.g., "acceptable". Popular Photography magazine has also published tests which showed manual focusing to be superior overall, and showed a surprising degree of inaccuracy in AF. And no, I don't know what issue it was anymore.
My own limited experience with AF is that Nikon is equal to manual focusing in good light and does well in the EV 0 range if I find it an area with good contrast. My experience with Canon has not been so good. It seems like like a "good enough" approach is taken, and have had images which showed inexact focusing even in good light (defining it as no less than open shade on a sunny day). Pressing the button multiple times improved results. I never used a Canon in really dim light so I don't how it does. A problem with either was the tendency of multiple focusing points to mean a lot of hunting or selecting away from the subject. I had a greater problem with the Canon consistently wanting to pick nearer objects, even with the subject centered. This has held true with a 10D I used recently.
This is singular, anecdotal evidence and should be taken as such.
The dinky viewfinders of crop-sensor digitals do make accurate manual focusing harder, compared to 35mm.