I can buy this argument. But in practice how much does it matter? It's not illegal to say "you may like this terrorist recruitment video", especially when its just a generic recommendation algorithm making that recommendation.

That “generic recommendation algorithm” is anything but. It’s proprietary and serves their interests and completely opaque. Who knows if it’s an algorithm or actually being hand picked?

It's not illegal but they're civilly liable for the harm they cause.

I don't think so. 230 protects them from liability for the video, and 1A protects them from liability for the recommendation. It would be hard to argue that a recommendation isn't speech.

How is this any more their speech than company employed moderators picking and choosing which content users can view or pinning posts?

That's certainly an interpretation of things, but it runs afoul of an issue where you'd force the Supreme Court to legislate acceptable types of content suggestion.

Is a forum showing you the most upvoted (reddit's /top) or the most recently commented (phpBB), or some combination (HN and reddit's /hot) and saying "you may like {{most popular thing}}" the site's speech or the user's? Is it only the site's speech if the recommendation is personalized? Or if the recommendation is a single item, and not a list?

The court's 230 interpretation today[*] seems to be that suggesting or promoting user-content isn't itself site content, which is fairly in-line with the original intent of the law (that showing, and even ordering user-content isn't site-content).

[*]: Which as others mentioned, is totally independent of this ruling.

I agree with that.

However... Isn't it a little bit of the same thing when a tech company says "This isn't a illegal in any way, but we're going to use our intentionally ambiguous TOS to remove it".

Aren't the things you say similar to the things you chose to prevent from being said?

No, because Section 230 specifically and extremely explicitly protects one of those things and not the other.