Producing original data vs. citing authoritative sources well

Producing original data vs. citing authoritative sources well

The question: to demonstrate expertise, is it more valuable to publish original research (surveys, tests, datasets) or to cite established authorities accurately and abundantly?

These serve different ends. Original research makes you a primary source others cite — the most durable position, and the one generative engines preferentially quote because it is non-substitutable. Strong citation makes you a reliable secondary source — useful, but inherently derivative and easier to replace.

The QRG's treatment of expertise rewards content that demonstrates command of a subject; original data is the strongest possible demonstration. Studies of which pages get cited in AI Overviews and large-language-model answers (e.g., analyses of citation patterns through 2024) repeatedly find statistics, named studies, and quotable specific claims over-represented — original numbers are citation magnets.

The tradeoff is cost and risk. Original research is expensive and can be wrong, exposing you to correction. Citation is cheap and safe but ceilings out at 'competent summary,' which is exactly what generative models now produce for free.

Caveat: bad original research is worse than good citation — a flawed survey actively damages trust. The advantage assumes methodological honesty.

What we still don't know: how reliably current systems distinguish genuine original research from fabricated or p-hacked numbers presented with the same surface confidence.