It's actually pretty funny to think about other AI scrapers ingesting this nonsense into the training data for future models, too, where the last line isn't enough to get the model to discard the earlier false text.