I was listening to an older episode of Lexicon Valley, pre-John McWhorter, in which a guest linguist was talking about research into current language changes. He suggested that the mail that had been receiving over the previous few years was a treasure trove, with better data than researchers were going to be able to get with their tape recorders and listening to popular media. The audience would be a group of people interested in language usage with some amateur training in it, just from listening to the show. They are spread into thousands of places with listening ears, including some obscurer corners and unusual backgrounds by chance, such as people who had moved here from other countries (including English-speaking ones) which would be logistically difficult for a researcher to find in any quantity by intent.
They are not a perfect filter, certainly, as they will tend to be drawn from the ranks of the more educated and wealthier, and they might hear things wrong or listen for the wrong things. But a linguist could easily work around that. It's crowdsourced information, and even if imperfect, very useful.