BS. Non-business flights are the most price-sensitive type of product there could possibly be.
Near 100% of people only apply a few very basic filters, e.g. no 12-hour layovers, doesn't depart at 5am, and then choose the cheapest. It's not some sophisticated "balance" that needs some advanced machine learning or personalization to accomplish. It's a few obvious filters that are the same for practically everyone.
I'd love to see partial dependance plots for each of the obvious filtering factors, vs how often people chose that flight. That's the best way to do this type of analysis.
Not to mention that many times the cheapest flight listed on these travel sites isn't actually available when you try to book it. Wonder how that may have skewed Google's analysis.