The problem is targeting and segmentation. If a hundred thousand people see evidence for X and a hundred thousand people see a debunking that always convinces y

The problem is targeting and segmentation. If a hundred thousand people see evidence for X and a hundred thousand people see a debunking that always convinces you of not-X, the debunking doesn't matter unless it's the same hundred thousand people. If the segment is the roughly 100M voters in America, you'll reach one in a thousand unless you've got the same targeting/segmentation filters.