[lhorie]

This is pretty simple to bypass, e.g. type this in the url:

javascript:d=document.createElement("div");d.innerHTML="<style>span:nth-child(odd) {display:none;}</style>";document.body.appendChild(d);

[nbush]

Nice! That's probably the easiest way. I guess you could scatter the spans a bit more randomly, but I couldn't find a way around JS completely.

[lhorie]

Scattering the spans more wouldn't do much... it's still pretty straightforward to gather the list of css classes used in the page and just piece together the innerText of their elements until you see a dictionary word (or even easier, a space).

In any case, there's always the impossible-to-defend-against attack: just hit PrintScreen on the keyboard, paste to MS-Paint and run the image through one of the many OCR websites that you can find via Google. Maybe even run it through MS-Word's autocorrect after that.