SEO news: June 2025
Google Keynote
Back in late May, the Google I/O 2025 conference introduced a number of innovations for users. In some ways, it even brought groundbreaking changes.
🔍AI mode and Deep Search
Google is testing a new feature called Deep Search in its AI mode. The “Expert Search” generates in-depth, professional, and cited reports. It uses hundreds of queries at once to produce complex summaries that would normally take an average user dozens of minutes, or even hours, to gather by browsing various websites.
🧠 Personal Context
The AI Mode can leverage personal data to tailor responses such as information from Gmail, browsing history, or connected devices - customized to your bookings, preferences, or travel plans. However, this personalized approach should only occur with the user’s prior consent.
📸 Live Search: real-time visual answers
The new Live Search brings interactive visual search. Point your phone at an object and the
📊 Custom Charts & Analysis: create custom charts
AI Mode can generate custom charts and analysis, such as sports or financial statistics, and visualize them to the user's requirements. All this is done based on information that Google has in its index from external sites. Graphs are created in "real-time".
🛍️ AI Shopping: intelligent shopping
Google is testing two additional user conveniences: agent shopping and try-on functionality.
- Agentic shopping - you select goods according to defined criteria, AI searches for products and buys them if you want. For example, if they are on sale. To complete the purchase, the user has to make a final confirmation by physically clicking "buy". Payments are via Google Pay.
- Try-on function - try on dozens of t-shirts virtually without having to go to the store. Just upload a photo of your figure.
🤖 Agentic Mode: AI as assistant
AI can book tickets, order services or fill out forms for you. So far it's in the testing phase in Labs and working with partners like Ticketmaster, StubHub, Resy and Vagaro.
AI mode diagram - how does it work?
🧠 1. State-based, context-based search model
Google AI Mode does not work with queries in isolation. It uses context - query history, location, device, and user behavior - and converts them into vector views to better understand long-term intent.
🔁 2. Query Fan-Out: Synthetic Queries
When you enter a query, AI Mode automatically generates dozens to hundreds of related queries into its database (query fan-out). It searches a wider range of relevant documents and resources, and then generates output based on this mix.
📚 3. Custom corpus and Multi-Stage Modeling
From these results, a custom corpus is built - a selection of documents that the model then processes using several specialized LLMs: a summarization model, a structured data extraction model, a matching model, and others.
🧩 4. Obtaining relevant passages
Instead of entire pages, AI Mode works with passage-level retrieval - it evaluates and selects specific sections of text in vector space that best match the query. Thus, the competition is not just between sites and pages, but also between the quality of text sections.
👥 5. Personalisation with help of "user embeddings"
AI Mode creates vector user profiles based on previous behavior (e.g., clicks, history, Gmail). Thus, two users with the same query may receive different answers tailored to their preferences.
🏗️ 6. Multiphase LLM processing and synthesis
The next phases involve models for creative content generation, fact validation, and multimedia creation, depending on the nature of the query. The goal is to ensure that the response is coherent, reliable, and relevant.
What does this mean for organic performance?
We are entering an era of truly valuable content for users. What worked in the past may no longer apply to conversational language models and answer-driven search engines. Creating structured and rich content that can be easily cited and utilized by large language models (LLMs) will become an essential part of achieving visibility and success in organic search results.
Summary
- Innovative Search: AI Mode delivers synthesized answers, multimodal content, agent services.
- Interventions in the visibility of information queries: content without clicks still builds authority.
- Google AI Mode brings a paradigm shift: Moving from classic full-text search to synthesis and personalization.
Recommendations
Visibility will need to be measured not only by traffic or impressions, but also by citations in AI outputs.
Optimize for topics, not keywords. Covering the user search intent - that is, the purpose behind a user’s query - will be crucial in content-focused SEO.
Keep originality, depth, and authority in mind. These long-standing quality criteria have always been important, but they are now gaining even greater significance.
Prepare the content for the separation of individual passages. The text should therefore be structured, factual and quotable.
Don't forget multimedia. It is advisable to use additional formats such as images, videos, etc.
Sources used
https://io.google/2025/explore/google-keynote-1
https://blog.google/technology/ai/io-2025-keynote/#generative-media
https://www.seroundtable.com/google-ai-mode-io-39444.html
https://ipullrank.com/how-ai-mode-works
https://developers.google.com/search/docs/appearance/ai-features