Creating headlines that are relevant to readers is a task that all print copy writers have faced over the past century. A large body of knowledge exist that can be applied. Much of this came from Sears in their early catalog development. Sears ran multiple headlines and graphics in controlled experimentation. Response levels were monitored to develop copy and graphic rules.
The cost of testing within the web environment is much cheaper and should be part of any web development plans. The simple way to do this is to systematically make changes to key page elements and determine how the rate of visitor responses changes using the site visitor tracking package provided by most server services. For example I plan on systematically testing various recipe descriptions to determine which descriptions generate the greatest percentage of visitor clicks to the specific recipe within the site. The key is not to vary more than one core element at a time and maintaining base behavior data to compare the changes with. This is as important as search engine optimization efforts for evolving a successful website.