That is a very generous way of thinking about (no doubt) very smart people.
There is a lot of precedent in using the mystery shopper technique to assess companies that aren't considered Human Subjects Research even if the interface to that company is a single human being.
Your analogy with mystery shopper doesn't hold the water as well since the researchers in such cases are just executing the primary function of the businesses in question. So while for the bigger sites you might be correct that the research is not a human subject research but it's still an abuse of service which in unethical without consent as well.