Bill Slawski was kind enough to point me over to a newly released patent application from several Googlers focusing on how paid ads might be served, including the order of the ad results and the inputs used to determine relevancy and scoring.

Here's a quick breakdown of the patent, as I don't have time to give it a full review:

1. How to accept ads and determine the concept associated with each including:
1. Using keyword targeting information
2. Determining similiarity between ads
3. Using 4 items to determine a score for ads
1. Ad performance information
2. Advertiser quality information (where do they get this?)
3. Ad price information
4. An Information Retrieval score (could this come from Google's normal index scoring of the sites?)
4. Relative preference figuring into scores
2. Triggering ads based on concepts using:
1. Concept vectors and the "dot product" of the concepts in the search request (basically a way of examining if the query is relevant to the ad)
2. Figuring historical performance information into the trigger to determine if the ad is worth showing
3. Measuring an ad against other similiar ads from the past to see if it is likely to be relevant
3. Methods for determing "concepts of requests" (by requests, I believe they mean search queries) including:
1. "no concept" requests
2. performance information based on "advertisement selection" information
3. performance information based on conversion information
4. Matching ads to documents based on extracted concepts (as well as search results using:
1. Relative preference for certain ads with certain concepts
2. Use of prior search query information to determine which ads to serve

The patent also says it will be:

• "increasing the score if the tracked performance information is above a threshold performance level" and
• "decreasing the score if the tracked performance information is below a threshold performance level"

I'd be on the lookout for new types of Adwords and Adsense services in the near future based on this document.