Yes, an LLM as any statistical model converges on the statistically most likely result, i.e. barfed out text, but what the value of this is? When references & citations are often fabricated/confabulated, I'd rather just occasionally groan at dumb WP articles and roll eyes at Reddit threads.
