At Google Search Central Stay Deep Dive Asia Pacific 2025, Cherry Prommawin and Gary Illyes led a session on how AI matches into Search.
They requested whether or not we want separate frameworks for Generative Engine Optimization (GEO) and Reply Engine Optimization (AEO).
Their insights counsel that GEO and AEO don’t require wholly new disciplines.
Picture taken by creator, Search Central Stay Deep Dive Asia Pacific, July 2025AI Options Are Simply Options
Cherry Prommawin defined that AI Mode, AI Overviews, Circle to Search, and Lens behave like featured snippets or data panels.
These options draw on the identical rating indicators and information sources as conventional Search.
All of them run on Google’s core indexing and rating engine with out requiring a standalone platform. Including an AI element is just a matter of introducing further interpretation layers.
Gary Illyes emphasised that each AI-driven instruments and basic Search companies share a single, unified infrastructure. This underlying infrastructure handles indexing, rating, and serving for all end result sorts.
AI Mode and AI Overviews are simply options of Search, and constructed on the identical Search infrastructure.
Deploying new AI capabilities means integrating further fashions into the identical system. Circle to Search and Lens merely add their query-understanding modules on prime.
Crawling
All of the AI Overviews and AI Mode options depend on the identical crawler that powers Googlebot. This crawler visits pages, follows hyperlinks, and gathers recent content material.
Gemini is handled as a separate system inside Google’s crawler ecosystem and makes use of its personal bots inside Google’s ecosystem to feed information into its fashions.
Indexing
In AI Search, the core indexing course of mirrors the strategies used for conventional search. Pages which have been crawled are analyzed and arranged into the index, then statistical fashions and BERT are utilized to refine that information.
These statistical fashions have been in use for greater than 20 years and had been first created to assist the “did you imply” characteristic and assist catch spam.
BERT provides a deeper understanding of pure language to the combination.
Picture taken by creator, Search Central Stay Deep Dive Asia Pacific, July 2025Serving
As soon as the index is constructed, the system should interpret every person question. It appears for cease phrases, identifies key phrases, and breaks the question into significant components.
The rating part then orders lots of of potential outcomes primarily based on numerous indicators. Completely different codecs, corresponding to textual content, photos, and video, carry totally different weightings.
RankBrain applies machine studying to regulate these indicators whereas MUM brings a multimodal, multitask method to understanding complicated queries and matching them with the absolute best solutions.
What This Means: Use The Similar Ideas From website positioning
Given the tight integration of AI options with normal Search, creating distinct GEO or AEO packages might duplicate current efforts.
As SEOs, we should always be capable of apply current optimization practices to each AI Search and “conventional” Search merchandise. Specializing in how AI enhancements match into present workflows lets groups leverage their experience.
Spreading assets to construct separate frameworks might pull consideration away from higher-impact duties.
Cherry Prommawin and Gary Illyes concluded their session by reinforcing that AI is one other characteristic within the Search product.
website positioning professionals can proceed to refine their methods utilizing the identical ideas that information conventional SEO.
Extra Assets:
Featured Picture taken by creator
