[multicast]

But isn't this a win situation for Google to a certain extent? Since it uses up the budget of the advertiser much faster. And the accuracy of filtering new revenue coming from ads as a company is already fairly limited in general. But maybe there are multiple reasons that Google really only wants to serve real humans to the ads of its clients.

[jsnell]

It's not a win. The fake clicks will not convert to sales, and the advertisers are seeing a lower ROI on their ads and will go and spend their budget elsewhere in the future. All ad networks will try to filter out as many fraudulent clicks as possible, because they are not optimizing for the maximum revenue today but for the revenue in the long run.

But yes, of course this is not just about filtering out fake clicks. The draft proposal lists a bunch of use cases, most of which have nothing to do with ads.

[multicast]

Interesting explanation, I totally agree on the click-per-pay part. But how would you track the benefit of ads with paying-per-impression? I know its less expensive, but according to the article paying per view seems to be a quite big part of the ad business.

[jsnell]

The article is just straight out wrong about "Google’s ad network charges per impression". The author clearly doesn't know anything about the area, made up some shit on how things could work, and just wrote it into their article with no fact checking.

You're right that attribution and measuring ROI is way harder and less precise for ads sold by impression than by click. That's why they're not the common form of advertising, especially on these kinds of ad networks. But for cases where the ads are per impression, the concerns about fraud would be exactly the same. It's not about a crawler accidentally generating impressions, it's about bots deliberately doing so.