But we know that isn't actually the case. The algorithms are designed to drive engagement, what they show you is a function of your behavior on the site, it's not an "endorsement" by the site creators, they're finely tuned machine learning systems optimized for ad dollars - that's it.
Once again, this is just another example of how the ad model incentivizes pernicious behavior. If these sites were under a subscription model they'd be happy to offer you the option to sort your feed with a purely chronological algorithm because they'd still be getting paid regardless of your engagement behavior.
There are also active efforts to remove “bad” content from such recommendations, which is effectively un-endorsing it.
You can take HN as an example of a recommendation service - while they remove things that are off topic, you'd have to argue that YC endorses whatever is in the #1 spot on HN.
HN has more active, hands on moderation than most social media sites, and there's a stronger argument that dang "endorses" the front page of HN than that Steve Huffman "endorses" the front page of reddit.
At the time, websites were mostly just hosts of content. The individual is posting on the website but otherwise the website is a tool.
This changes with recommendation algorithms. Nearly all social media is based on some kind of recommendation algorithm. Should that be covered by 230? It could be argued that it starts to get closer to an endorsement of certain content (and indeed, some of that recommended stuff is ads.)