But in this case it's the site choosing to share the search term with the ad network -- and www.medicinenet.com/gout/article.htm (my first result for gout) has a pretty good idea that you searched for gout. It's also a rare enough term (this is my first time ever typing it) that any calculation on statistically improbable phrases will know I searched for "gout", not "article".

In most cases the "leakage" is pretty minor from search engine to page, the big leak is from page to ad network.

You are affected if you see ads, using https, using noscript, blocking ads and erasing most of your cookies you are somewhat not vulnerable.

Using google the first link that came with 'gout' was wikipedia.

Still as an advertising demographic "someone researching gout for whatever reason" is more valuable to target than a viewer you know nothing about -- even if half of them are researching the Henry VIII, the other half have a new condition and they need to buy something if only they knew what.

"But in this case it's the site choosing to share the search term with the ad network"

in the same way that most users probably aren't aware of this, most publishers probably aren't aware as well. i, for one, never considered that the referrer might be passed along to the ad networks i use on my sites. it simply never occurred to me. i definitely didn't choose to share that information with them.

agreed. I guess I left out my suggestion for what the problem really is:

If you are concerned about Ad Networks having so much data on you, clear your cookies and block cookies from them. Then they will never be able to string together more than one piece of data.

You're only bumping the problem from a per-browser (cookie) aggregation to per a machine or household (IP address) aggregation.

Let's not slip into a false dichotomy. We don't need to argue about whether to block cookies OR block search engine leakage.