Introduction to AI in Search
At the Google Search Central Live Deep Dive Asia Pacific 2025, Cherry Prommawin and Gary Illyes led a session on how AI fits into Search. They discussed whether separate frameworks are needed for Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO). Their insights suggest that GEO and AEO do not require new disciplines.
AI Features Are Just Features
Cherry Prommawin explained that AI Mode, AI Overviews, Circle to Search, and Lens behave like featured snippets or knowledge panels. These features draw on the same ranking signals and data sources as traditional Search. They all run on Google’s core indexing and ranking engine without requiring a standalone platform. Adding an AI component is simply a matter of introducing extra interpretation layers. Gary Illyes emphasized that both AI-driven tools and classic Search services share a single, unified infrastructure.
AI Mode and AI Overviews are just features of Search, and built on the same Search infrastructure.
Deploying new AI capabilities means integrating additional models into the same system. Circle to Search and Lens simply add their query-understanding modules on top.
Crawling
All the AI Overviews and AI Mode features rely on the same crawler that powers Googlebot. This crawler visits pages, follows links, and gathers fresh content. Gemini is treated as a separate system within Google’s crawler ecosystem and uses its own bots within Google’s ecosystem to feed data into its models.
Indexing
In AI Search, the core indexing process mirrors the methods used for traditional search. Pages that have been crawled are analyzed and organized into the index, then statistical models and BERT are applied to refine that data. These statistical models have been in use for more than 20 years and were first created to support the “did you mean” feature and help catch spam. BERT adds a deeper understanding of natural language to the mix.
Serving
Once the index is built, the system must interpret each user query. It looks for stop words, identifies key terms, and breaks the query into meaningful parts. The ranking phase then orders hundreds of potential results based on various signals. Different formats, such as text, images, and video, carry different weightings. RankBrain applies machine learning to adjust those signals while MUM brings a multimodal, multitask approach to understanding complex queries and matching them with the best possible answers.
What This Means: Use The Same Principles From SEO
Given the tight integration of AI features with standard Search, creating distinct GEO or AEO programs may duplicate existing efforts. As SEOs, we should be able to apply existing optimization practices to both AI Search and “traditional” Search products. Focusing on how AI enhancements fit into current workflows lets teams leverage their expertise. Spreading resources to build separate frameworks could pull attention away from higher-impact tasks.
Conclusion
Cherry Prommawin and Gary Illyes concluded their session by reinforcing that AI is another feature in the Search product. SEO professionals can continue to refine their strategies using the same principles that guide traditional search engine optimization. By understanding how AI features work and how they integrate with traditional Search, we can optimize our content and improve our online presence.