Trump ad ... click Comcast ... click Ambulance chasing lawyer ... click click click
So that would be better than random - a plugin that clicks the ads for things you don't like or aren't interested in.
I also think that prices are extremely inflated.
Even stuff which shows like hundred people a day costs you 8 dollar per click. No amount of selling will you get back the money.
Imho something is wrong there...
I'd guess on the keywords where law firms are spending hundreds per click, you're dealing with malpractice and personal injury cases. Those cases are likely to be done on contingent and can have pretty huge upsides. I'd think a well-run firm would have some idea of the expected value of those types of cases and can spend accordingly to acquire clients. Considering how targeted search ads are compared to how these firms normally advertise -- local TV, radio, billboards, busses, etc. the costs probably aren't radically different than they're used to spending on advertising.
If that were true these ads would just go away, or cost less. They are making money.
Depending on the exchange, you can sometimes see how much the DSP/client paid for the impression by looking at the win notification URL. Typically there will be a URL query parameter that has been filled in by the exchange, like `BID_CPM=4.00`, and when your browser requests this URL (which is owned by the DSP/client), your browser is effectively notifying the winner of the auction that they won and how much it cost. However, ad exchanges are increasingly sending these fields encrypted to protect against manipulation/fraud.
If you really want to manipulate the amount of money that advertisers pay to show you ads, look at what ad botnets like the one in this article do. The PDF above only briefly mentions it, but one of the things they do to increase their perceived value to DSPs and clients is they make their browsers visit high value advertisers websites (e.g. they might visit a shopping website and put items in their cart to make it look like they were strongly considering purchasing the items). Later, when you visit a publisher site, this will cause the bidders' machine learning systems to predict a much higher value for you than they otherwise would have, and submit higher bids as a result.
(Have they come up with "no fly" lists over there yet?)