In the past, I was responsible for analytics around a decision to autoplay or not autoplay videos on a media site. The site did both mobile and desktop videos.

    * voluntary customer feedback was universally negative
    * bounce rate decreased substantially (around 40%)
    * total interactions with the page increased almost 3x (clicks on significant  UX elements such as like/subscibe/additional video plays)
    * interactions for those using some form of adblock stayed the same (presumably these people also had an autoplay off plugin)

In summary, what users say they like and what they actually like are wildly different. The average user seemed to engage more when we removed the burden of first interaction.

*We did eventually kill autoplay, but only so we could have a consistent experience between YouTube and non YouTube videos (YouTube took efforts to stop counting autoplayed videos for advertising purposes).