But when a modal pops up or I get redirected, or your ad floats over your content and tries to follow my scrolling not only are you not going to sell to me, but I'm going to leave your website.

I had to watch youtube without an adblocker recently for the first time pretty much ever.

The ads were not only extremely irrelevant. They were extremely sexist, strongly offensive and discriminating. Without exception (note I'm not in the US, so these probably aren't the same you're seeing!). I was incredibly surprised that most of the population watching youtube was exposed to this daily but not completely outrage.

You would have no way of knowing if the image was an ad or part of the article content since they were all served through the same thing. No silly subdomains like ad.domain.com either.

Still quite feasible to detect by size and styling.

I wish web developers had more respect for their audience. If I say, "No ads," I really mean "No ads." Your site is not a special snowflake where I suddenly don't mind seeing advertisements.

I'm also not sure what you mean by "respect for their audience". If I put an ad on a website I didn't ask your opinion and frankly I don't care what you think. Unless you wanna pay my bills. If what you're saying is I should detect the existence of ad block and actually proactively stop writing the code for said ad into the website.....well... that's just silly.

But that only defeats simple adblocks that blacklist domains. It's theoretically possible to make an adblock that detects ads through machine learning and removes them. Doesn't matter where they are hosted.

But advertisers can run scripts to detect modifying the web page and stop it. So adblock would have to move to having a headless browser that renders the page, and then copies it and removes the ads.

Then the advertisers would have to try to trick the machine learning itself. Adversarial machine learning is a thing, with neural networks trained to fool other neural networks. And images that are highly optimized to exploit subtle defects in other neural networks.

As ML ad detection improves, the only way for ads to get past it will be to make ads that look more and more like real content, even to the human eye. Perhaps in the limit, ads and content will become completely indistinguishable.

This is already happening. Further, embedding paid for content in the middle of original content at the server side is also already happening.

> As ML ad detection improves, the only way for ads to get past it will be to make ads that look more and more like real content, even to the human eye. Perhaps in the limit, ads and content will become completely indistinguishable.

This is also already happening. Brand placement deals are nothing new to advertising and they are basically table stakes to most content purveyors.

The real question isn't if those things are going to happen, its how those sorts of deals will handle the analytics side of the equation. How will you prove that your instagram feed is the best place for the yoga pants provider to spend their marketing budget?

I don't know whether this is a deliberate reference, but there is South Park episode playing with this idea.

https://en.wikipedia.org/wiki/Sponsored_Content_%28South_Par...

They are already doing it: Google AMP, Facebook Instant Pages.