Mostly the same foundation, with a different surface. AI search leans heavily on the same underlying signals classic ranking has always rewarded (authority, relevance, and trust), but it adds real divergence at the surface layer, where extractability, direct-answer phrasing, and claim-level clarity matter more than they ever did for a blue link. So it is neither totally different nor exactly the same, and the useful answer maps the two layers rather than picking a pole. This is fast-moving, so treat the map as the current understanding and worth confirming.
The shared foundation is the larger part. AI systems retrieve and trust the same kinds of pages classic search surfaces: credible sources, relevant to the query, with the authority signals that distinguish reliable information from noise. The work that builds rankings (earning links and mentions, covering a topic well, being trustworthy) is the same work that gets a page into the pool an AI draws from. If you have built that foundation, you are not starting over for AI.
The divergence is at the surface, in how the answer is presented. Classic ranking can reward a page for being the best overall result even when its specific answer is diffuse. AI selection rewards the page that states the answer in a clean, self-contained, liftable chunk, phrases it directly, and makes the specific claim easy to extract and attribute. Two pages with the same authority can split here: the one that hands over a crisp answer gets cited, the one that buries it does not.
That is the whole map. Underneath, the signals overlap so much that good SEO is most of good AI optimization. On top, AI adds a presentation layer (extractability, direct phrasing, claim clarity) that classic ranking did not weight as heavily.
Keep the shared foundation you have already built (authority, relevance, trust) and add the AI-specific surface on top: make your best pages state their answers in clean, direct, self-contained sentences that a system can lift and attribute. Do not rebuild, just resurface.