There's also a subtype that deliberately circumvents this trick. There is a type of ad fraudster that rips off of, e.g. native German-speaking writers, runs their work through machine translation, and publishes the English output as original.

I encountered several examples on some technical subs on Reddit (not yet on HN). It took a disproportionate amount of effort to unmask even one -- the method that ended up succeeding was to guess which technical terms could be idempotent under translation, and (&&) some together until the result set is small enough. (It's harder to reverse translate, because unless you're an expert translator, you probably don't know what the source language was).

I've only seen a handful of these, but because of how difficult it is to detect, I'd speculate there could be a sizeable population in the wild. The writing is technically correct and non-suspicious, because it's written by a human expert in another language. It strongly resists reverse Google searches. And it resists social unmasking, because social groups who speak different languages tend to have distance between them.

It's a clever evil.