I used to work at an advertising company that worked this way, doing lead gen. In some verticals there's nothing better, and as long as your own metrics are good you can figure out how much a traffic source is worth. If a traffic source goes bad and starts being 50% spam, no problem, your automatic system lowers its bids by 50% and you're fine.

The thing is, only a very small fraction of the ads ecosystem works this way. Coke wants to show ads to people to position their brand, not to get people to buy soda over the internet, which means they're not able to measure the efficacy of channels on a per-publisher basis and figure out if they're being cheated. Brand advertisers typically use third-party fraud detection services like Moat to inspect the environment their ads are running in and see if they look legitimate.

Then there's a large range of people doing things where you could measure the effect of channels, but the site just isn't that sophisticated. Some pizza place somewhere buys some ads, they don't have the tech staff to measure conversions and optimize their channels, instead they trust the ad network to make sure they're not being cheated.

Even sophisticated performance advertisers rely on ad network fraud detection, just indirectly. Yes, you run your ad campaign, and you bid what the traffic is worth to you based on your measurements. But the existence of that traffic stream comes from fraud detection. Imagine some publishers on an ad network start spamming it, and the ad network isn't able to detect the fraud. The money from advertisers now isn't going to publishers in proportion to the value of their traffic, but instead in proportion to how much they're cheating. The publishers that were the source of your good traffic go out of business, and the ad network collapses because it doesn't have valuable traffic to send to anyone.

I recognize that other advertisers may have different goals, such as for large, brand-centric advertising campaigns. I'm not an expert on those so can't offer an informed opinion on any specifics. From a casual observer's point of view, I don't see how they're any worse off than the classic "everyone is wasting half of their ad budget, they just don't know which half" of the print/broadcast era. As you say, at least with online ad networks there can be some indication of where the ads are being shown, allowing a level of independent scrutiny by the advertiser if they wish.

Try looking at it from the publisher's perspective. If their ad network can't detect fraud then of all the money the network brings in an increasing proportion will go to fraudulent publishers. Honest publishers will make less and less money and eventually give up on the ad network.

This does seem to assume that ad networks are the main or only way to support sites that rely on ads for their funding. I'm not sure why we should make that assumption, though.

We're talking about a built-in feature to safely and transparently include ad content from a third party in your own site. It seems possible that advertisers could set up standardised content to work with that feature. Then the kinds of hosts who make enough from advertising to be worth keeping it anyway could either delegate to an advertiser's own server or self-host. Payment could be either a direct transfer or use some new, simpler and more transparent service just to handle that aspect.