Google Search Central APAC 2025: what really counts in SEO practice

Introduction

A lot from the Google Search Central Live Deep Dive APAC 2025 sounds like a detail update. In everyday project work it is not. Google made clear that there is no separate strategy just for AI features: AI has long been in every step of search, from crawling to output. That answers, for Google Search, a question many still discuss as a field of its own.

I have seen the consequences in projects for months. How Google handles “Discovered – currently not indexed,” what weight crawl budget gets, how AI works into every step: that already decides performance today. I go through the points that count day to day.

Content and ranking

Content for people, not for machines

The core message was unambiguous. Content written for people works. Google’s machine-learning models train on high-quality, human-created text — Gary Illyes said it on stage in exactly those terms. The more natural a text, the better Google classifies it. Pages with real, readable language beat those still working with keyword stuffing.

Direct and indirect ranking signals

Google weighs direct and indirect signals together. Direct are the words on the page and their position in the main content. Indirect are links and mentions from other pages. Whoever works only on on-page optimization leaves half the equation lying.

E-E-A-T is an explanatory principle, not a ranking factor

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is not a direct ranking factor. It explains how Google assesses quality, nothing more. There is no “E-E-A-T metric” that flows into ranking. The principle shows through measurable things: the author profile, source citations, evidence of real expertise, trust signals.

PageRank lives on

PageRank is still in use. The public metric of old is history; internally a further-developed variant runs. Link structure and a website’s authority therefore remain decisive. For practice this means: internal linking deserves more attention again than it was lately granted.

AI and search

Why no separate strategy just for AI features is needed

AI is not an add-on to search. It sits in all processes. AI Overviews and AI Mode build on the same search infrastructure as the blue links. Whoever does solid SEO thereby makes their content fit for the AI features too. Google puts it this way itself: no separate framework for GEO or AEO is needed; whoever does classic SEO well is qualified for the AI features anyway. One caveat belongs to it, and Google did not spell it out on this stage: the sentence holds for Google Search. Answer engines like ChatGPT or Perplexity draw their sources by their own rules.

AI is in every step of the search process

AI runs through the whole process. In crawling it steers the schedules by website performance and user behavior. In language understanding, BERT and related models capture context and intent. Spam is caught by SpamBrain. In ranking, RankBrain and MUM take part.

Query fan-out and grounding

In “query fan-out,” Google breaks a complex query into several sub-queries and thereby gathers information comprehensively. In “grounding,” the system checks generated texts against the index to reduce hallucinations. For content planning, a clear rule follows: a text should carry the individual sub-question just as well as the overarching topic.

Quality beats origin

Google assesses content by quality, usefulness and reliability; origin plays no role. AI tools may therefore co-write. Human review afterward remains mandatory.

AI Overviews and higher engagement

AI Overviews are, per Google, the biggest change to search in 20 years. Users search more often and are more satisfied. “AI Mode” takes on complex queries for deeper research. Whoever wants to appear there thinks in whole questions rather than single keywords.

Technical SEO

”Discovered – currently not indexed”

The status means: Google wanted to crawl the URL but would have overloaded the website with it, and postponed it. Instead of getting annoyed about unindexed pages, work on server performance and crawling efficiency.

”Crawled – currently not indexed”

If a page was crawled but not indexed, it is down to content quality. The instruction is accordingly direct: improve content. In practice that means a content audit that uncovers duplicate content, thin pages and missing substance of your own.

Crawl budget: server errors cost, 404s do not

5XX server errors burn crawl budget, 4XX errors do not. But 4XX affect the planning and prioritization of crawling. The focus therefore belongs on server errors. How the HTTP status codes steer crawling in detail I described in the piece on crawl budget.

Deduplication and canonicalization

The canonical URL for duplicates is chosen by Google via machine learning. Redirects, content similarity and rel=canonical flow into the decision. Important: canonical statements are recommendations, not commands. A single signal is rarely enough. Let internal linking, URL structure and clearly distinguishable content work together.

Schema and structured data

Too much micro-markup bloats pages and still does not move ranking directly. The benefit lies elsewhere: schema clarifies entity relationships and feeds the LLM features. Use it deliberately for important entities and features, not blanket.

Geotargeting

For regional content, Google recommends ccTLD, hreflang and localized content per version. For international e-commerce projects this point weighs especially heavily.

User behavior and new tools

Generation Z

The 18-to-24-year-olds are, per Google, the fastest-growing group of searchers. They search more directly, more visually, and expect fast, precise answers.

Visual search: Google Lens grows 65 percent a year

Google Lens grows 65 percent a year, with over a hundred billion searches in 2025 alone. “Circle to Search” is available on more than 250 million Android devices. Among Gen Z, around 10 percent of searches already start through such AI entry points instead of the search box. Whoever wants to be found there optimizes product images seriously: high-quality photos, descriptive alt text, structured data for images.

Search becomes more visual

Search becomes more visual and interactive. Content should therefore sit in the text and the image and carry several forms of interaction.

Practical implementation for your 2025 SEO strategy

Rethink content strategy

Write for people first, but structure so that AI systems understand the content: clear language, clean structure, complete answers. Serve several formats, that is classic search, AI Overviews, visual search and voice search. And think in entities and their relationships, not just in keywords.

Technical priorities

For server performance one thing counts first: eliminate 5XX errors, secure with monitoring and failover. Align crawling efficiency with business goals — important pages easily crawlable, quick to load. Use structured data deliberately with Schema.org, without bloating the pages.

International strategy

International projects stand on ccTLD, hreflang and localized content. To this belongs a real cultural adaptation that goes beyond mere translation. And local authority: regional E-E-A-T signals through partnerships, mentions and backlinks.

Monitoring

Keep an eye on the crawling status, that is “Discovered/Crawled – currently not indexed” in Search Console. Track your presence in AI Overviews. And check visibility in visual search, keyword Google Lens.

Trends and forecasts

AI integration becomes the norm. SEO managers learn to work with AI systems without losing sight of the human user. How strongly AI Overviews spread Google said on stage itself: the biggest change to search in 20 years.

Visual search moves into everyday use. The 65 percent growth at Google Lens is only the beginning, and e-commerce has to catch up on its image optimization. Personalization runs increasingly through AI. And the technical bar rises: crawl budget, server stability and structured data demand more technical understanding than before.

Frequently asked questions (FAQ)

Do I need a separate strategy just for AI features? No. In Google Search, AI already sits in all processes. Optimize your existing SEO so it carries for classic search and AI features: high-quality, clearly structured content with complete answers. For answer engines like ChatGPT or Perplexity, separate rules apply; Google said nothing about that on this stage.

Why do my pages appear as “Discovered – currently not indexed”? Google wanted to crawl but postponed it, so as not to overload the website. That is a signal to work on server performance and crawling efficiency: improve load times, reduce server errors, sharpen internal link structure.

Is E-E-A-T still important if it is not a direct ranking factor? Yes. It is an explanatory principle and shows through author profile, source citations and trust signals. Show real expertise instead of ticking off a checklist.

How do I prepare for visual search? Google Lens grows 65 percent a year, and among Gen Z around 10 percent of searches start through Circle to Search or similar AI entry points. Optimize images systematically, with descriptive alt text and structured data. In e-commerce, product images from several angles belong to it.

Should I use AI-generated content? Google assesses not the origin but quality, usefulness and reliability. Use AI for speed and human review for quality.

How important is PageRank today? It runs on, as an internal, further-developed variant. Link structure and authority remain important. Work on clean internal linking and on high-quality external links.

My conclusion

Back to the start: a lot from APAC 2025 sounds like detail but is not. Google rewards readable content written for people, holds the technical infrastructure to be decisive, and continues to count E-E-A-T as an explanatory frame, not a lever. AI is not a special case here, it sits in every step.

For your work this means something concrete. Open Search Console this week, filter to “Discovered – currently not indexed” and “Crawled – currently not indexed” and separate the two causes: one is a server and crawling matter, the other a quality matter. Then check, for your five most important pages, whether the central content really stands there as HTML text and does not first load via JavaScript. Two checks, one afternoon. After that you know where to start.