Specifically NetChoice argued that personalized feeds based on user data were protected due to first person speech. This went to supreme court and supreme court agreed. Now precedent is set by the highest court that those feeds are "expressive product". It doesn't make sense, but that's how the law works - by trying to define as best as possible the things in gray areas.
And they probably didn't think through how this particular argument could affect other areas of their business.
It absolutely makes sense. What NetChoice held was that the curation aspect of algorithmic feeds makes the weighting approach equivalent to the speech of the platforms and therefore when courts evaluated challenges to government imposed regulation, they had to perform standard First Amendment analysis to determine if the contested regulation passed muster.
Importantly, this does not mean that before the Third Circuit decision platforms could just curate any which way they want and government couldn't regulate at all -- the mandatory removal regime around CSAM content is a great example of government regulating speech and forcing platforms to comply.
The Third Circuit decision, in a nutshell, is telling the platforms that they can't have their cake and eat it too. If they want to claim that their algorithmic feeds are speech that is protected from most government regulation, they can't simultaneously claim that these same algorithmic feeds are mere passive vessels for the speech of third parties. If that were the case, then their algorithms would enjoy no 1A protection from government regulation. (The content itself would still have 1A protection based on the rights of the creators, but the curation/ranking/privileging aspect would not).
I misunderstood the Supreme Court ruling that it hinged on personalization per user of algorithms and thought it made a distinction between editorial decisions that show to everyone vs individual users. I thought that part didn’t make sense. I see now it’s really the third circuit ruling that interpreted the user customization part as editorial decisions, not excluding the non-per user algorithms.
This ruling is a natural consequence of the NetChoice ruling. Social media companies can't have it both ways.
> If that were the case, then their algorithms would enjoy no 1A protection from government regulation.
Well, the companies can still probably claim some 1st Amendment protections for their recommendation algorithms (for example, a law banning algorithmic political bias would be unconstitutional). All this ruling does is strip away the safe harbour protections, which weren't derived from the 1A in the first place.
And they probably didn't think through how this particular argument could affect other areas of their business.