LLMs Re‑centralize Expert Gatekeepers

Large language models, when coupled with reliable retrieval, tend to produce consensus‑aligned, evidence‑based summaries that can shift influence away from viral social‑media actors and back toward expert sources and curated institutions. That re‑centralization could change who sets political narratives, how journalists source stories, and which institutions wield epistemic power. — If true, this reverses a core consequence of social media — decentralised populist amplification — and would reshape debates about trust, censorship, platform policy, and AI governance.