The real (not fictional) watermarks have a particular purpose -- to identify content after it was processed or modified, or included into other content. In other words, after some technology or creativity was applied to it. One can, say, scale , then convert a video clip from mpeg2 to sorensen3, and the result still will have the original watermark recognizable. Or, say, a clip or an image, or sound used as a part of something else, and the result still can be found to contain the watermark. The primary reason is to look for unauthorized use of watermarked elements in published material made by others.

One does not need a watermark to track the distribution of unmodified material on public-accessible sites -- to check if something in particular is there, still has to download everything, and comparing the file with the original is easier than the check for watermark.

End user, who just receives the material and watches it, gets nothing visible, and, obviously, can not be tracked. The only way to track the user is to have his viewer software modified (I would call it "sabotaged") that it looks for watermarks and reports it to the authors (or to some Random Goons Association of America). Without the second part the user is just as anonymous as without the watermark itself.

Of course, no one would voluntarily and knowingly install a sabotaged viewer if the alternative is present, but this is where "Random Goons" are doing their best to make the alternatives hard or impossible to get, so authors who merely want to track the end users might get a windfall of those efforts.

Also I have to add that everyrhing above is how things would work if watermarks were as reliable as the originators of the idea wanted them to be. AFAIK, many pretty trivial transformations of content remove the watermarks.