Ad fraud is a big problem in the digital ecosystem. A simple, straightforward example of ad fraud is a publisher who wants to make more money. They get paid through exchanges on CPM basis (or CPC for adsense), so they write a computer program that keeps reloading their page, and sometimes clicking. This increases the views - hopefully without decreasing the CPM too much - resulting in more yield. This is the most lucrative in the video space where CPMs are high and quality inventory is low.
You might think it'd be simple to catch a bot, and the above bot would be. But bots are written to avoid being caught. Some of the more sophisticated bots infect computers around the world, meaning the traffic isn't traceable to some number of computers, then they load websites using a hidden version of a browser - one never viewable by the user - and tries to mimic user behavior, including scrolling, clicking on random things, hijacking that user's cookies (which might have real purchases in their history) etc. This is when it gets really hard to detect.
There are companies like Integral Ad Science (aka IAS), White Ops, and Google (via its acquisition of spider.io) that try to identify these bots. It's really hard - given what I described in the second paragraph, how would anyone know for sure what is or isn't a bot? I personally have no idea, but I also don't think about it all day. We partner with some of these companies to detect fraud, but it's only so reliable. We've generally see it be directionally accurate when applied to suspicious publishers on our platform.
As part of our value of "take pride in your craft" we try to have a high quality platform with as little fraud as possible. We're definitely not immune to fraud, but one of the things that's interesting about native ads is that they're different across different publishers, meaning it's a completely unique HTML placement. It's hard for a generic bot to find and click on our ads (though not impossible). A bot could certainly be programmed to find and click on ads on one site, but it's a lot hard to do it generally.