It’s amazing what people “just know” without having any basis in fact.
Math (statistics) will tell you that a random sample size of less than 2,000 is necessary to be 99% sure that your data is within +/- 3 percentage points of accurately describing the entire population—whether that population is a million or a billion.
Of course, the study may not have been designed properly, you’d have to examine their methodology. And even if the design is sound, it may not have been executed correctly.
But Google didn’t take issue with the study, so they apparently didn’t have a problem with the results, even if you do.