As with any security measure the limitations of a security measure depends on the particular implementation scenario. Digital watermarking of text can be hard to implement if you are on the middle of the publication chain and editing is done to the document before publication. This can happen with both encoding on text or encoding on white spaces.
When you control all the publication steps up to the actual distribution of the document is when the measures can be more effective.
If the attacker suspects there is a possible watermark, depending on the particular method, it may be possible to remove the watermark. Again, the success of the technique depends on the particular security scenario that you are expecting