In a recent conversation with Jesse Dwyer of Perplexity, key insights emerged about how SEO professionals should adapt to the rise of AI search optimization. His perspective clarifies not only what’s changing—but what still matters—in this new era of search.
First, Dwyer emphasized that personalization is transforming search results. “It’s no longer a zero-sum game,” he explained. Two users entering the same query can now receive different answers—especially when using AI tools like Perplexity or ChatGPT that incorporate personal memory into the context window. This means visibility is no longer about ranking in a single, universal list of results.
However, traditional SEO still plays a foundational role. Perplexity, for example, uses a form of PageRank—a link-based system—to determine which content is eligible for retrieval. So while the output may vary per user, your content must first be deemed relevant and authoritative by classic ranking signals.
Dwyer then outlined a crucial technical distinction: whole-document versus sub-document processing. In traditional search, engines index and rank entire webpages. AI tools built on this model—like ChatGPT’s web search—essentially run a standard search, pull the top 10–50 pages, and ask the language model to summarize them. This approach is often called Generative Engine Optimization (GEO), but it’s still rooted in conventional SEO.
By contrast, true AI search optimization—or Answer Engine Optimization (AEO)—relies on sub-document indexing. Instead of storing full pages, the system breaks content into tiny semantic units: snippets of 2–4 words, converted into numerical vectors via transformer models. When a user searches, the engine retrieves up to 130,000 tokens (roughly 26,000 snippets) to completely fill the AI’s context window.
Why does this matter? Because saturating the context window with relevant fragments reduces hallucination. The model isn’t left guessing—it’s fed precise, grounded information. As Dwyer put it, this shifts the AI from “creative generator” to “accurate answer engine.”
Moreover, personal context further shapes results. The AI can use known user preferences, past interactions, or location to refine which snippets populate the window. This explains why two people get different answers to the same question—even when drawing from the same knowledge base.
The real competitive edge, Dwyer noted, lies in how companies retrieve and weight those snippets. Techniques like query reformulation, dynamic compute allocation, and proprietary ranking models help surface the most relevant fragments. For publishers, this means content quality, semantic clarity, and topical depth matter more than ever—not just keyword placement.
In summary, AI search optimization requires a dual focus: maintain strong traditional SEO fundamentals (especially backlinks and site authority) while also structuring content for granular relevance. Write clearly. Cover topics comprehensively. Avoid fluff. Because in the age of AEO, your sentences—not just your pages—may be the unit that wins visibility.
As the industry moves toward sub-document systems, understanding this shift isn’t optional. It’s essential for any publisher or SEO aiming to thrive in AI-driven search.
READ: Thailand Approves $3.1B in Data Centre Projects in 2026
