How to prepare for Google's AI Mode and LLM Search Discovery

Google has stated that AI mode is the future of Google Search. Do you know how it works or how to appear? This session will outline how to appear in both AI Mode and other LLM search platforms. We’ll walk you through how to set up reporting and audience research to determine how much these platforms are impacting you today. Then we’ll walk through how these Large Language Models (LLMs) generate their answers, find and provide references to those answers, and present them to users. We’ll also walk through free and paid tools to help you pivot and be found in this new search landscape.

Key Takeaways
- What Google AI Mode is and how it's going to impact your nonprofit
- How to create a plan to appear in Google AI Mode and measure your impact
- What other LLM search platforms are on the horizon, and how to create a plan to be visible in their search results.

Transcript

1. How to prepare for Google's AI Mode and LLM Search Discovery Katherine Watier Ong AI for Nonprofits & Advocacy Summit - July 31 2025
2. Katherine Watier Ong WO Strategies LLC • Built first website in 1994. • Worked in-house at nonprofits for 11 years. • Lost sites due to too much traffic (EWG.org for 4 days, but still 1.2 million visits up from 300K) • Set & led social media customer service strategy for NY Health Obamacare Exchange. • Developed and ran the 360 digital marketing strategy for HealthIT.gov. • Helped Cancer.gov grow from 1.5 million to nearly 5 million searches/mo. • Wrote the first study on consumer adoption of wearable computers. (2003)
3. Brands Katherine Watier Ong has worked with
4. 1 The New Digital Ocean 2 The Evolving Digital Landscape • Text embeddings everywhere ◦ AI Overviews ◦ Google's future for AI Mode • Do you need to focus on LLM Search? 3 The Blue Ocean Framework • Eliminate • Reduce • Raise • Create 4 Measuring Success Rework your KPIs. What can we *now* measure? Today’s Agenda • Zero click search & social • Loss of accurate marketing data • Discovery behavior has changed
5. Navigating The New Digital Ocean
6. The Seismic Shift in Digital Discovery 1. AI search drives almost zero clicks. 2. We have lost most traditional web analytics. 3. In 2025, people don’t start with Google if they have questions, they start Googling once they have found a solution.
7. Search, Talk and Listen with AI Mode Now 100M monthly users In US, UK and India
8. Google’s AI Powered Web Guide
9. Google's future is Agentic Search "The future of Google Search, a search that goes beyond information to intelligence" - Google
10. How have the AI Overviews Impacting Search?
11. 44% of nonprofit website visits come from organic search. Most websites rely on free traffic Source
12. All info queries have AIOs = no clicks Source
13. When an AIO appears, it dominates • If a Featured Snippet and AI Overview co- appears, it takes up 67-76% of the mobile screen. • They are also now appearing before Discover posts. Source https://pewrsr.ch/4lIqbsM
14. Health, Society and Science show most Source Desktop Mobile
15. What Happens when an AOI Appears?
16. Jan 2025 - Organic CTR w/AI .64% Source
17. AI Overview Growth Currently 2B monthly users. These rankings change hourly.
18. Impressions Up, Clicks Down (50%+)
19. The Leftovers/Long Tail is *Tiny* Big brands and topics are winning. “The Long Tail/not branded queries (<11 searches/month) is only 3.6% of demand. Google has become a place people go *after* they discover a need rather than a demand-creation or even demand-nudging platform.” Rand Fishkin, CEO Sparktoro
20. • Only 34% start with search • 32% start with social media 2025 users don’t look for ideas via Search April 2024 data
21. Gen Z (%) Millennials (%) 44% of Millennials use Google for discovery, compared to 29% of Gen Z. Gen Z and Millennials: Largely Social To Discover New Things Source
22. Social platforms reward zero click content
23. Social is ranking in Search “Perspectives” • Study.com • Quora.com • YouTube • Facebook • Instagram • TikTok • Reddit Keywords in captions!
24. Old Paradigm: Fighting for Clicks • Biggest marketing budget • Longest -established website • Site with the most backlinks • Ranking #1 • Keywords • Getting clicks from search
25. New Paradigm: Fighting for Citation & Brand Recall • Semantic triples (brand + topic) consistent across ecosystem • Brand searches • Embedding vectors • Measuring success by brand recall and AI citations
26. 2025: The Loss of Marketing Data
27. Organic Search Tracking Issues - GA4 • You can not track absolutes • Impossible to know how many users you have unless you have a login • You can’t measure the bounce rate • AI Search, many mobile visits, and others are lumped into “direct” • You should be tracking percentages and sharing trends
28. Organic Search Tracking Issues • People decline cookies and remove UTMs • “Not set” are people who keep your website open in a tab. • 67% of US adults block cookies or the browser does (Safari, Firefox, Brave)
29. AI Overviews & AI Mode and these that doesn’t pass a referrer and are measured as “Direct”: ***also: radio, TV ads, magazines/newspapers, in-store announcements, word of mouth, packaging, and billboards. TikTok WhatsApp Instagram DMs Facebook Messenger Discord Reddit Slack LinkedIn Messages Apple iOS X DMs Dark Search & Social - Direct Mastodon
30. • AI Mode & AI Overview traffic is not segmentable in GSC • AI Mode & Overview visits are lumped together with Featured Snippets and People Also Ask. • You can purchase SEO tools (ZipTie.dev) to track AI Overview rankings. Not Trackable: AI Overviews and AI Mode
31. Most LLM Referrals End up as Direct
32. AI search estimated to overtake traditional search by 2027.
33. The Evolving Digital Landscape
34. Users are delegating discovery These AI systems: • Take a questions • Evaluate the landscape • And offer a single output - an answer, product or decision. They evaluate. They summarise. They decide. -Jono Anderson
35. It’s not about clicks anymore
36. How Google’s AI Overviews & Mode works
37. AI Overview
38. Powered by Gemini 2.5 pro: • Uses "query fan-out" to combine results from multiple related searches (from the Google index) into one comprehensive answer • The references applied to that answer are from Google's index and selected because they have a block of text that has high cosine similarity between the text and the topic. • It can accept multimodal inputs and creates multimodal outputs • Is personalized based on the user’s text embedding • Now includes ads. Google's AI Mode: Radical Transformation
39. What is Query Fan Out? • A method of breaking down the initial query (or prompt) into multiple related sub-queries that are concurrently searched by an AI platform. • The results from all those searches are synthesized into a comprehensive response to the original question (or the original search query). Prompt in Gemini
40. Here’s Google’s visual for how it works:
41. Query Fan out - Gemini (plugin)
42. Personalized AI Overview User Location Search History Device Data Your Google account User Demographics User Embeddings for each Google user
43. It’s encouraging longer queries
44. Text Embeddings & Cosine Similarity Measure the cosine similarity between a topic and passage on any URL by using AI Overviews Visualizer. A cosine similarity of 1 means identical, and zero means not related.
45. How is this measured? 1. AI converts the word and the passage to vector embeddings (a list of numbers that captures the semantic context). 2. Then it calculates the distance between the two vectors to determine if the keyword and topic are semantically related.
46. Built off Traditional SEO Rankings - AIO
47. AI Mode is Trained on Google’s Quality Signals
48. Content Optimization =a Credible Fragment • Your fragment is credible and has citations, quotes, and stats. • No more than 5-6 citations per fragment • Cosine similarity is close between topic and fragment • H1/H2 in header. • Organize your content into interlinked topical clusters. AI systems use Reciprocal Rank Fusion (RRF). • Get our guide here on writing Fragments/Fraggles
49. Ahrefs research on elements that correlate with AI Overview visibility. Wikipedia, YouTube, Reddit and .gov show 3x more often in AIOs then regular results.
50. Google only Ranks Helpful Content Guide around how to provide helpful content lists questions to ask. • Quality content • Written from a palace of expertise, with author info clearly displayed (bio) • Good UX (Google watches us on the Chrome browser) • Content created for people with a purpose that will help them with their goal. WHY are you creating the content. • Check with this custom GPT
51. Does LLM Search need to be on my radar?
52. Google Search processes 373 times more searches than ChatGPT. LLM search send 96% less traffic to news sites and blogs than traditional search engines.
53. AI Search Growth Source
54. The Blue Ocean Framework for Nonprofits
55. Blue Ocean Framework 1 Eliminate • Information easily answered by AI. 2 Reduce • Click KPIs, rank tracking • Unhelpful content • Siloed teams 3 Raise • Broad topical semantic coverage • Brand signals & mentions. 4 Create • Plan to be found in LLM data. • Local expertise content. • Unique data products
56. What you need to STOP doing
57. • Stop chasing broad, high-volume keywords • AI Overviews increasingly dominate informational queries • Focus on niche, specific terms instead ELIMINATE: Content that AI can answer
58. Google is indexing less. Focus on authoritative, in-depth zero click content that showcase your nonprofit's brand and perspective. Google Quality Rater Guidelines Penalize: • "Scaled content abuse" • Content created with "little effort or originality" • AI-generated content that offers no added value ELIMINATE: Content for Content's Sake
59. REDUCE: Complex, Jargon-Filled Language Over half of American adults read below a sixth-grade level. AI prioritizes clear, easily understandable content, as do users.
60. REDUCE: Siloed Digital Marketing Website Social Media Email Marketing Public Relations Community Engagement
61. Old KPIs: • Rankings • Clicks • Organic traffic volume New KPIs: • Impressions • Share of voice for target topics • Direct conversions Acknowledge that a dip in website traffic is inevitable as users get answers directly from AI. REDUCE: Change your KPIs
62. RAISE: Become a Trusted AI Source
63. Add unique insights & schema Improve readability Increase EEAT & entities Create fragments, headers & measure similarity Identify persona, queries & then fan out Optimization for AI Overviews & AI Mode
64. Google Understands Your Journey “...to deeply understand … how interests can develop over time as familiarity and expertise grow.” -Google Source
65. Create a Relevant Content Fragment • Clearly answers the search intent in a concise, readable way. • The paragraph has high cosine similarity to the keyword target • Answer the query fan out and use expected entities. • Has a nested sequential relevant header Helpful Tools: • MarketBrew's AI Overviews Visualizer • Custom GPT for measuring cosine similarity • Source
66. AI systems use entities to understand the relationships between different pieces of information and the user's intent. Your next steps: • Create topic clusters around the entities. • Create a central pillar page with a complete overview & multiple cluster pages with strategic internal links. • Your text should have the entities they expect. Use this GPT. Cover the expected entities
67. Citations Link to authoritative external sources that support your claims Expert Quotes Include direct quotes from recognized authorities in your field Statistics Present relevant data points from credible research Aim for no more than 5-6 citations per content fragment to maintain focus Credibility is Key
68. RAISE: Brand Signals
69. Give Google your Entity Data 69 Might result in a Knowledge Panel, or a voice result, or rich snippets. How? • Be a brand • Use structured markup (schema.org) on your website (especially sameas) • Be in Wikipedia/Wikidata • Link to external entity references. • Google’s advice • **Also confirmed in Google algo leak
70. Wikipedia & Wikidata Claim and improve your pages on these authoritative sources. Use a coach with Wikipedia. Knowledge Panels Claim and enhance your Google Knowledge Panel. Entity Hub Use sameas, nested under organization schema on your about page to link to these external references. Can also be used as a part of persona markup on author pages. Brand Entity Optimization
71. Example Sameas Markup
72. Check: Schema Use this Schema Markup Testing tool to see if schema is set up correctly.
73. CREATE: Improve what they know about you
74. What do LLMs know about me?
75. Nonprofit Brand Name Relevant Topic Keywords Improved AI Visibility Co-occurrences (brand + topic)and context influence LLM mentions Co-mentions and Context
76. Query Fan- Out
77. Query Fan out for ChatGPT Options for seeing queries used in fan out in ChatGPT: • ChatGPT search query extractor (bookmarklet) • Better GPT (Chrome plugin) • ChatGPT Path (Chrome plugin)
78. AI Search: Built off SEO Rankings
79. • High-quality blogs and news sites • YouTube videos and descriptions • Authoritative publications • Q&A sites like Reddit or Quora Increase Brand Mentions
80. Your brand + the LLM You need to check your brand entities across all of the LLM chats. Prompts to try: • What is X brand known for? • What topics is [brand] authoritative on? • What brand is high authority in the [topic] or [service]? • Who is known for X topic targeted at X people? A more elaborate process to find the questions that people ask that are associated with your brand.
81. INCREASE: LLM friendly content
82. You provide your semantic triples and brand to the AI system, and hope it still remembers you when it creates a summary. Semantic Triples
83. Train LLMs on your Semantic Triple 1. Benchmark your current visibility. Prompt LLMs to see if you’re recommended. 2. Track your AI Overviews, and rankings that could be mentioned in AI Overviews with help. 3. Create YOUR semantic triple and make sure its mentioned in LLM training data. 4. Audit your content for entities coverage, readability, and optimized passages. 5. Fix any inconsistencies in LLM by testing changes to pages mentioned. 6. Add sameas markup on your site and make sure it mentions your triple. 7. Add other schema markup: author, organization, FAQ formats 8. Get cited in journalism used in the training data
84. The new maintenance - LLM auditing Periodically prompt LLM with queries about your brand: ➢ spot-check how you’re being described, omitted, or misrepresented. Reinforce weak signals: ➢ Identify outdated or low-trust sources and upgrade them. Reclaim neglected listings. Submit corrections where needed. Patch reputation vulnerabilities: ➢ Identify and fix inconsistencies in how your brand is described across surfaces. Address negative discourse early, before it spreads or hardens.
85. • Create comprehensive resource hubs to signal deep authority to AI models • Link them together internally • Video formats (YouTube) are shown regularly in AI Overviews/mode CREATE: Multi-Format Content Hubs
86. INCREASE: LLMs’ awareness of your unique data
87. Train LLMs with Your Unique Data • Publish original data in crawlable formats (comparison with tables, FAQs). Add Schema markup • Create interactive tools/unique insights • Share with public training sources (Wikidata, Data Commons, Github)
88. INCREASE: Feed LLMs your Local Expertise
89. Become the trusted local source for LLMs • Optimize your Google My Business profile. Make sure it mentions your triple. • Clean up your Name, Address, Phone number (NAP) citations (software needed). • Encourage reviews and respond to them. • Develop pages that target local searches (if relevant) • Address real community questions with SEO-friendly Q&A • Add Local Business schema markup, and Event markup on events. Test the markup.
90. INCREASE: Build AI-Powered Engagement Tools
91. CREATE: Custom GPTs and LLM Pages • Custom-branded GPTs in ChatGPT • Pages within platforms like Perplexity • Train on your organization's specific content • Provide direct answers about your mission and programs • Include donation and volunteer pathways • If you have a lot of data, consider an LLM partnership
92. INCREASE: Make sure LLMs bots can see your content
93. 1 Fast Loading <500 ms 2 HTML Content (no JavaScript) 3 Crawlability Allow crawling with your robots.txt 4 Indexability Make sure Google/Bing can index your pages (used for RAG) Technical SEO for AI Crawlers
94. Your new Robots.txt For an updated list of AI agents, check Dark Visitors.
95. What you *can* measure
96. Old KPIs • Rankings • Clicks • Organic traffic volume • Bounce rate Shifting Key Performance Indicators New KPIs • Impressions across topic clusters • Share of voice • Direct conversions • Brand mentions • AI crawl rate (log files) • Citation frequency in AI AI Overviews -daily rank changes & 70% of ranking pages changing every 2-3 months
97. GA4 - Tracking LLM Referrals • One off exploration report -> • Create a custom channel report: ◦ Video walk-through & explanations from Kickpoint here • Looker Studio report lets you compare your LLM traffic against organic traffic
98. • AI Mode & Overview visits are lumped together with Featured Snippets and People Also Ask. Here’s how to set up reporting for that. • The only tool that tracks logged in Google users: ZipTie.dev. • AI Mode tracking is available in DragonMetrics. No filterable data for AI Overviews and AI Mode
99. Measure LLM Rankings • ZipTie.dev • OtterlyAI • BrandRank.ai • SpyGPT • RankScale.ai • Profound (enterprise) • Waikay • Full spreadsheet of options here.
100. New Marketing Measurement Options • Brand impressions (Buzzsumo) • Clickstream data tools (SimilarWeb, Sparktoro, Audiense) • Brand lift studies (pre and post campaign) • LLM rank tracking (see our list), AI Overview tracking (Ziptie.Dev) • Share of voice tracking across platforms (Keyhole, Sprout Social) • Output from SEO tests (like using Search Pilot)
101. Run SEO Tests • Test if your change creates lift. • Monitor how and why your competitors are appearing and try to reverse engineer. • Here’s a great overview on how to run SEO tests. • Tools: ◦ https://seotesting.com/ ◦ https://www.searchpilot.com/
102. Implement Strategy Apply AI optimization techniques to key content Measure Results Track new KPIs and monitor AI inclusion Analyze Performance Identify what's working and what isn't Adapt Approach Refine strategy based on findings Continuous Testing and Adaptation
103. Key Takeaways • Google's AI Overviews drastically change information discovery, and have dropped CTR to 1%. • Nonprofits need to integrate their social and SEO work to maintain organic impressions. • Content creation has to change for cosine similarity, passage creation, and broad, interconnected topical coverage. • Marketing KPIs need to change. Most of this is unmeasurable. • LLM search is growing and nonprofits should benchmark and create a plan for growth now before the market is more saturated.
104. Additional Resources
105. Resources • AI checklist from Aleyda Solis • How AI Mode works - Mike King, iPullRank • This custom GPT: Can it rank for an AI Overview? • Our guide on how to rank for AI Overviews • To learn more about embeddings - see the free course here.
106. Bonus Items • Creating a social media strategy that supports Google discovery. • SEO Writing Checklist • Searcher’s Intent tutorial and checklist • Featured Snippets writing guide and checklist • Image Optimization Checklist • Fraggles Writing Checklist • Guide for improving CTR • Optimizing the SERP Snippet Generated with Gemini Advanced
107. A SEO writer’s checklist • Does this match intent (content type, format & copy)? • If an AI Overview appears, does your passage match the topic? • Have I explained why my content matters to the topic? • Did I back my claims up? • Is every sentence adding value? • Have I given a next step? • Did I review the copy based on EEAT? • Am I encouraging the click from search?
108. KATHERINE WATIER ONG 202-930-1744 katherine@wostrategies.com www.linkedin.com/in/katherinewatier www.WOStrategies.com Daily SEO tips via Alexa