That specific scenario doesn't make economic sense. There's too much variability for the insurance company to make a good decision. Consider that a health teacher may assign students to research different common ailments, as part of a lesson to understand the available tools; a science teacher may bring up the story of H. pylori as a digestive disorder which was misunderstood for a very long time, and the student wants to learn more; the teenager may be helping a stressed out friend; the teenager might be interested in going into medicine and researches common diseases; and so on.
To start with, every insurance company develops a statistical model of each and every customer. Non-smokers get lower rates than smokers. Life insurance depends on age. That's how insurance companies must work.
That article says that some insurance companies use more than just risk factors in order to determine the price, but also information like market power (or lack thereof) for your market segment. For example, poor or immigrant people may not have the time or knowledge to look for alternative companies, so an insurance company may offer an initially low price to get lock-in, then raise it after a few years to be above the price needed for actuarial reasons.
It doesn't talk about online browsing data, and it doesn't talk about personalized factors other than the expected aggregate demographic modeling.