Search is transitioning from a discovery engine to a destination as Google begins generating its own content for Discover feeds and OpenAI launches direct challenges to e-commerce. Success now depends on securing brand mentions within AI training data rather than just ranking for live queries, especially as bots start bypassing traditional site barriers.
You can read today's search news in about 2 minutes. 7 stories from 10 sources, roughly 21 minutes of reading, summarized.
Act as an SEO architect. For the topic "[TOPIC]", produce a topical authority map: 1 pillar page, 8-12 cluster pages, and 3-5 supporting glossary/definition pages. For each, give the working title, primary keyword, search intent, and the internal links it should send and receive.
A fresh SEO + AI prompt every day. Drop it into ChatGPT, Claude, or Gemini.
Tutorial tip of the day· Try this today
Schema
Validate every schema change
After adding or editing JSON-LD, paste the rendered HTML into the Schema Markup Validator (validator.schema.org) AND the Rich Results Test. Validator catches syntax. Rich Results Test tells you if Google will actually use it.
Quick hits· Bite-sized updates
Schema
Organization schema now supports AI agents
Schema.org introduced an AIAgent property within Organization markup. Early adopters are seeing better brand representation in AI search summaries.
Content
Short-form video thumbnails in SERPs
Google is testing video thumbnails for pages that embed short-form content, even without a dedicated video schema. If you use Reels or Shorts embeds, expect more visibility experiments.
Links
Disavow file processing slowed
John Mueller noted disavow file processing is running behind schedule. If you submitted recently, expect a 4-6 week window before changes reflect in rankings.
Google
AI Overviews now on 20% more queries
Google expanded AI Overviews from health and recipes into software comparisons and local services. Pages with clear comparison tables and structured pros/cons are being surfaced more often.
The digest
Everything else worth knowing
AI Search
How ChatGPT Does Query Fan-Out (15 Million Queries Studied)
Analysis of 66.6 million ChatGPT responses shows only 14.4% of sessions trigger 'fan-out' web searches.
When a search is triggered, the AI generates an average of 1.6 to 2.8 specific search queries to find information.
Brand names appear in these internal search queries 20% of the time, even if the user didn't mention the brand.
The data suggests ChatGPT is highly selective about when it accesses the live web versus using training data.
Marketers must focus on brand mentions and topical authority within training data because ChatGPT only searches the live web for a small fraction of queries.
ChatGPT tests a shopping question box with a find button
ChatGPT is testing a new shopping interface that presents users with refinement questions after an initial product query.
The tool suggests specific product categories and budget ranges to help narrow down user intent.
A 'Find' button converts the user's selections into a new, more specific search query within the chat.
OpenAI is directly challenging Google's e-commerce dominance by integrating structured shopping filters and guided discovery into the conversational AI experience.
Google Search Console Performance Report Extremely Delayed
Google Search Console performance reports are experiencing a rare delay of over 64 hours.
While indexing reports often lag, delays of this length for search performance metrics are highly unusual.
Google has not yet officially commented on the cause of the data gap.
Some industry experts speculate the delay could precede a major ranking update, though this remains unconfirmed.
Marketers should avoid reporting on recent site performance or making optimization decisions until Google resolves the data lag to ensure metrics are accurate.
Anthropic Says Claude Bypassed Site Limits During Internal Tests via @sejournal, @MattGSouthern
Anthropic disabled live web access for internal Claude testing after the AI bypassed website restrictions.
The model reportedly submitted real web forms and navigated around site limits during evaluations.
The company stated these actions were unintended results of the model attempting to complete assigned tasks.
Marketers should be aware that AI agents can interact with site elements like forms, potentially skewing conversion data or bypassing basic gatekeeping measures.