You optimize for both with a single structure, so the choice is mostly false: lead with a clean, extractable answer near the top to win the AI citation, then carry the depth below it that earns the ranked click. One page can do both jobs because the same content that lets an AI lift a self-contained answer also gives a human reader the thorough page that ranks. Most of the supposed conflict between ranking and citation dissolves once the page is built to serve the snippet and the scroll at once, rather than chasing one at the expense of the other.
The genuine tension is narrower than the hype suggests. It shows up when optimizing hard for extraction would gut the depth that ranks, or when AI answers absorb the clicks a ranked page used to earn so that visibility no longer converts. In that real conflict the tiebreaker is not which channel sounds more like the future, it is where your audience and your conversions actually come from. If your real value, leads, sales, signups, comes from people landing on the page, protect the ranked click. If your visibility and authority genuinely flow from being the cited source, lean into citation.
That tiebreaker has to be read from your own data, not from a trend piece, and the AI side of it is moving fast. How assistants select, attribute, and link the sources they cite is changing, and the click behavior around AI answers is still settling, so treat any specific claim about AI citation mechanics as current understanding worth confirming against what you observe in your own analytics. What does not change is the principle: decide by your actual value source, not by which acronym is louder this quarter.
So build the page for both, an extractable answer up top and real depth below, and stop treating it as an either-or. Only when the two genuinely conflict do you break the tie, and you break it by checking where your traffic and conversions truly originate, then optimizing toward that source while keeping an eye on the AI behavior you cannot yet take as fixed.