Search News You Can Use - SEO Podcast with Marie Haynes

Did you know Google "re-ranks" search results using AI that is trained on user data and quality rater scores? In this video, I break down revelations from the DOJ trial, including the "Glue" system, RankEmbed BERT, and why user satisfaction is now the most critical factor for SEO.

00:00 - Intro: Why user satisfaction is the most important ranking factor.
00:49 - Google’s Quality Raters Guidelines & Helpful Content documentation.
01:36 - Marie’s Theory: Traditional algorithms vs. AI re-ranking systems.
02:35 - Why Google says "Don’t chunk your content" (Danny Sullivan).
03:00 - Vector Search & Embeddings: How AI retrieval actually works.
05:40 - The DOJ vs. Google Trial: Leaks on user-side data.
07:44 - The "Glue" System (NavBoost): How Google tracks clicks & hovers.
09:03 - "Instant Glue": How search adapts to new intent (Nice vs. nice).
10:20 - RankEmbed BERT: The deep learning model Google hid.
12:21 - How AI re-ranks the top search results.
13:03 - BERT Training: Wikipedia, Books Corpus... and romance novels?
15:18 - The 3 Deep Learning Models: RankBrain, DeepRank, RankEmbed.
17:40 - RankBrain: Why it only re-ranks the top 20-30 results.
18:47 - The Helpful Content System: A machine learning classifier.
21:18 - March 2024 Core Update: The evolution of helpfulness signals.
23:00 - Frozen vs. Retrained Google: How algorithms are tested.
26:01 - Why Google runs 17,000+ live traffic experiments.
27:16 - Does Google use Chrome data for rankings? (DOJ evidence).
28:32 - The "AI Mountain": Why AI-generated traffic often crashes.
29:30 - How to optimize for user satisfaction (beyond keywords).
31:00 - Liz Reid on AI Overviews: Why users click for depth.

In this episode, Marie Haynes discusses Google's Universal Commerce Protocol (UCP) and its transformative potential for online shopping. UCP allows consumers to purchase products directly from search results and AI interactions, streamlining the shopping experience. Haynes explains how this protocol works alongside Google's Business Agent, which enables businesses to interact with customers through AI agents trained on their specific information. She emphasizes the importance of having accurate data in the Google Merchant Center to ensure effective communication and sales through these new AI interfaces.

00:00 Introduction to Google's Universal Commerce Protocol
01:30 Understanding Business Agents and Their Role
05:50 How UCP Works for E-Commerce
12:01 The Future of AI in Shopping
15:50 Addressing Viewer Questions on UCP
22:04 Current Implementations of UCP
27:04 Potential Downsides and Considerations for Businesses
30:09 Looking Ahead: AI and the Future of Shopping

The future of search isn't search; it's agents. Dr. Marie Haynes tackles the biggest shift facing the web: agentic AI. This episode explores the new wave of agentic browsers like OpenAI's Atlas and Perplexity, the massive implications of Walmart partnering with OpenAI, and the legal battles already beginning. Dr. Haynes maps out a near future where home robots order products without ever seeing a website. What does an SEO or business owner do when the user interface is gone? This is a must-listen for anyone building a career on the web.

And OH YEAH, this episode was edited by an app that I created in AI Studio's build feature. It actually worked!

Comment here in Marie's community: https://community.mariehaynes.com/posts/drafts/the-future-of-seo-agentic-ai-new-podcast-episode

It was an honour to attend Google I/O in person.
Search is changing radically! Google called it a new era of Search with Gemini 2.5 at its core.

This podcast episode was recorded live.

Stay up to date on AI with my newsletter: https://mariehaynes.com/newsletter
Immediate updates on important AI and SEO news in my community, the Search Bar: https://community.mariehaynes.com/spaces/11879397/feed

0:00 – 0:51 Welcome & recap of Google I/O invitation, outline of the live episode’s agenda
0:51 – 1:55 Screen-share setup; meeting Sundar Pichai and Sergey Brin anecdotes
2:40 – 4:15 What AI Mode is, why Google may replace “classic” search, and the query fan-out patent
4:15 – 6:02 How clicks train AI Mode; demo showing 59 source sites and the result carousel
7:03 – 9:14 Ethics of Google using web content; E-E-A-T impact on carousel inclusion
9:55 – 12:22 Accuracy, hallucinations, feedback buttons, and why original content matters
14:00 – 17:21 Freshness signals, library analogy for originality, Gemini Live phone demo
17:21 – 27:55 Gemini AR glasses test, personal-context search & privacy implications
32:21 – 35:22 Gemini in Chrome: on-page AI assistance, summarisation, e-commerce concerns
35:22 – 40:47 Project Mariner: browser agents that automate multitab workflows; job impact
40:47 – 43:59 Agent Mode & agent-to-agent checkout; custom charts/graphs in AI Mode
43:59 – 50:52 Deep Search (multi-query research) and uploading your own files for reports
50:52 – 54:42 E-commerce upgrades: “Try it On” virtual fitting, price-tracking, “Buy for Me” with Google Pay, Wing drone delivery
54:42 – 57:00 Gemini V3 multimodal video creation, Flow editor, AI content production costs
57:00 – 60:00 Google’s universal AI assistant vision, coming hardware, and industry upheaval
60:00 – 61:00 Sergey Brin on the web’s unpredictability ten years out; reflections on accelerating change
61:00 – 62:27 Dario Amodei’s forecast of 20 % unemployment from AI; balancing fear with opportunity
62:27 – 63:50 Learning-curve discussion: experiment daily, let Gemini/ChatGPT tutor you step by step
63:50 – 66:00 Question on falling organic clicks: focus on conversions, cite Will Reynolds’ data, report revenue over raw traffic
66:00 – 67:12 Who gets credit—SEO or CRO? Rebranding SEO as AIO/GEO and outlining the evolving practitioner role
67:12 – 68:32 Brand building beyond Wikipedia: podcasts, social presence, authoritative mentions; example of podcast mention boosting AI-overview visibility
68:32 – 69:20 LLMs.txt debate: no major model honors it yet, so implementing it is likely wasted effort
69:20 – 70:10 Server-log analysis of AI bots: seeing ByteDance and others, limited actionable value beyond selective blocking
70:10 – 71:10 Google Agent Space: 100-seat minimum, use cases for large agencies (onboarding, HR, process agents)
71:10 – 72:10 MCP for e-commerce customization tools; preparing your site for future agent-to-agent APIs
72:10 – 73:20 Speculation that many third-party agent frameworks will fade once Google/OpenAI integrate agents natively
73:20 – 74:10 Community learning benefits; SEO skillset as problem-solving edge in the AI era
74:10 – 75:25 Overview of the Search Bar Pro community, tutorials, “happy hour” discussions, member advantages
75:25 – 76:50 Coping with upheaval: lamplighter analogy, choosing optimism, hunting for new opportunities
76:50 – 78:00 Closing remarks: upcoming talks on BCIs and agents
78:00 – 87:08 Final chat exchanges, farewells, and session wrap-up

Google's Agent2Agent (A2A) protocol is here, and it's set to change everything. Announced at Cloud Next 2025, this new standard allows AI agents to talk directly to each other.

What does this mean? An internet powered by communicating AI, radical changes for businesses using tools like Agentspace and the Customer Engagement Suite, and a potential upheaval for traditional SEO.

I share thoughts about A2A, the Agent Development Kit (ADK), Agent Cards, and why the future of the web might look very different, very soon.

More, including links to all of the resources mentioned in my blog post:
https://www.mariehaynes.com/googles-agent2agent-protocol-will-radically-change-the-web/

Subscribe to my newsletter for all of the interesting news as AI changes our world:
https://mariehaynes.com/newsletter/