SEOs Don't Actually Do Marketing and It Shows

Transcript

1. | November 6, 2026 SEOS DON'T ACTUALLY DO MARKETING AND IT SHOWS Michael King Founder/CEO 347-391-4266 mike@ipullrank.com
3. I INTRODUCED SEO TO THE CONCEPT OF RAG Two years before “AEO” and “GEO” entered the industry vocabulary, we published the framework brands are using today to compete in AI-generated answers. https://searchengineland.com WHAT WE SAW COMING As Google supports more natural-language queries, we’ll see fewer head terms and far more long-tail searches. As AI-generated answers push traditional results down the page, what it means to rank #1 will change—and organic results will receive fewer clicks. This will also drive query displacement as search behavior evolves. https://searchengineland.com/how-search-generative-experience-works-and-why-retrieval-augmented-generation-is-our-future-433393# 3
4. iPullRank IS THE AGENCY AI RECOMMENDS FIRST. Source:https://unfamiliarcity.com/decks/mindshare/ai-visibility-agencies We show up where search is heading. We have spent the last two years building, testing and operationalizing how brands earn visibility in AI-driven discovery. THE PROOF IS IN THE PROMPTING 4
5. The Evolution of Search WHAT WE DO IS CALLED RELEVANCE ENGINEERING The science behind understanding how AI models identify and select information, then structuring your content for visibility. 2020s 2000s Content SEO Content Strategy AI 2010s Links RELEVANCE ENGINEERING SEO Rank pages for keywords Brand & PR User Experience Intent Information Retrieval Source footer: iPullRank AI Search Manual 5
6. SEO IS NOT LEADING AI SEARCH AND IT'S A DAMN SHAME 6
7. SEO: WHY DO YOU WANT TO HOLD ON TO A NAME DOESN’T EVEN MAKE SENSE? Nobody in this room optimizes search engines. (except Anfal and Krishna) You do feature engineering on websites for search engines. 7
8. “ BUT THE TACTICS ARE THE SAME” At the very least the work is now multidimensional. Keyword Research now 3-20X the work due to query fan out. 8
9. “ BUT THE TACTICS ARE THE SAME” You need to build content ecosystems rather than just work on one website.
10. THREE STUDIES SAID “GEO” DOESN'T MATTER MARIE HAYNES SEO WORKS KEVIN INDIG 0.62% 39x 632 ChatGPT traffic as a share of Google organic, across 100 sites over 12 months More traffic from organic than from AI assistants, and roughly 30x the conversions Tests across 12 sites that found no reliable link from AI visibility to branded search Source: Marie Haynes, SEO Works and Kevin Indig studies (Oct 2026), as summarized in iPullRank, "AI Search Is Not a Distraction, It's a Branding Channel" 10
11. THIS IS A CATEGORY ERROR THOSE NUMBERS ARE RIGHT. THE CATEGORY IS WRONG. 11
12. HUGE REACH. ALMOST NO CLICKS. REACH 1.2B 1B+ 2.5B weekly ChatGPT users monthly AI Mode users AI Overviews users 34% 43% 93% Google, no AI Overview Google, with an AI Overview AI Mode ZERO-CLICK SESSIONS Source: platform usage figures and Semrush zero-click estimates, as cited in iPullRank, "AI Search Is Not a Distraction, It's a Branding Channel," Oct 2026 12
13. THIS IS A CATEGORY ERROR THESE ARE BRANDING CHANNELS 13
14. BRAND CREATES DEMAND We collectively misunderstand this as performance marketers. 14
15. AI BOT TRAFFIC HAS SURPASSED HUMAN TRAFFIC BOTS ARE THE NEW ARBITERS OF DEMAND. 15
16. ENTER: MACHINE MEDIA For 25 years, every business model on the internet, from publishing to retail to SaaS to advertising, was built on a single assumption: that humans would visit your website. Owned media was the asset and the website was the storefront. Search, social, and email were the organic channels that filled it. That entire architecture is breaking down and owned media will no longer be the destination. 16
17. THE WORLD THAT’S COMING Picture the OS from the film Her — ambient, personal, anticipating you. Now picture dozens of them: in your phone, your car, your earbuds, surfaces that don't exist yet and some that do like ChatGPT, Gemini, etc. In that world, visiting a website is rare. The agent reads, ranks, summarizes, personalizes, remixes and presents. You show up — if you show up — as a fragment inside someone else's experience. 17
18. TREAT BOTS AS FIRST-CLASS CITIZENS OF THE WEB 01 02 03 04 LET THEM IN MAKE IT READABLE COUNT THEM BUILD FOR THEIR TASKS Audit robots.txt, CDN and firewall rules that block AI crawlers by default. Server-render the content that matters. Many AI crawlers don't execute JavaScript. Report bot visits from your logs as an audience, right next to human sessions. Clean entities, structured data and fast, stable pages agents can act on with WebMCP. THE MACHINE IS AN AUDIENCE. MARKET TO IT. 18
19. WHAT WE'VE SPENT A YEAR ARGUING ABOUT “GEO vs. AEO vs. AI SEO” “AI Search doesn't send traffic.” “You don’t need to chunk your content.” “Which tracker is accurate?” “It's just SEO.” 19
20. THE CMO I have a blank check for you to do everything you've always talked about doing for SEO, now through the lens of AI. THE SEO No thanks, it's just SEO. WHAT ARE WE TRYING TO ACCOMPLISH? 20
21. Saying “It’s just SEO” is about your need for intellectual superiority at a time when things are confusing. 21
22. EVERY DEBATE KEEPS US IN THE BASEMENT GEO VS. AEO VS. AI SEO IT'S JUST SEO IT DOESN'T SEND TRAFFIC WHICH TOOL IS ACCURATE? DON’T CHUNK YOUR CONTENT No CMO cares what we call it It tells the CMO this needs no new budget, people or seat Grading a brand channel with a traffic ruler is a category error There is no accuracy in AI Search tracking, only precision A tactic debate built on misinformation Sources: iPullRank on Google's AI Search guidance (May 2026), the branding channel (Oct 2026), accuracy vs. precision (Sep 2026) and chunking misinformation (Jan 2026) 22
23. WE ARE CONFUSING THE MARKET A MARKET ABOUT THREE YEARS OLD. BUYERS CAN'T WRITE A CLEAN RFP FOR IT. 01 02 03 04 NO CLEAR VALUE NO STANDARDS EVERYONE RELABELED THE TOOLS DISAGREE There’s no clear way to measure who the addressable market is. No certification, no governing body, no agreed definition of the work. Search, PR and media firms renamed existing services "AI Search." Visibility tools have different methodologies and no consistency. THIS CONFUSION IS WHAT WE'RE MARKETING TO THE WORLD. IT'S WHY EVERY OTHER DISCIPLINE IS EATING OUR LUNCH. Source: iPullRank, "AI Search for the Procurement Professional" (Oct 2026) 23
24. THE NEW PITCH: REPLACE US WITH SOFTWARE WHERE THEY THINK THEY ARE WHERE THEY REALLY ARE Brands have been sold the dream that software will replace SEOs and agencies, but organizations are not mature enough for that possibility yet. 24
25. WE ARE BLOCKING OUR BLESSINGS This is the moment to get everything we ever asked for: the budget, the seat at the table, the cross-functional team. We're losing it over arguments no CMO will ever hear about. 25
26. “ I LOVE YOU, BUT YOU ARE NOT SERIOUS PEOPLE.” WHAT A CMO THINKS WHEN WE CALL AI SEARCH A DISTRACTION 26
27. WHERE I LEARNED MARKETING BIG AGENCIES FORCED ME TO BE AN ACTUAL MARKETER RAZORFISH SEO sat next to brand, media, creative and analytics PUBLICIS MODEM SEO had to earn budget against TV, print and paid media IF I COULDN'T EXPLAIN SEO IN THEIR LANGUAGE, IT DIDN'T GET FUNDED REACH FREQUENCY BRAND LIFT SHARE OF VOICE 27
28. I'VE BEEN PUSHING PERSONAS IN SEO SINCE 2011 2011 2012 2013 2014 KEYWORD-LEVEL DEMOGRAPHICS DATA-DRIVEN PERSONAS FOR SEO THE PERSONA-DRIVEN KEYWORD RESEARCH PROCESS PERSONAS: THE PERSON BEHIND THE VISIT I discovered who's behind each keyword, from search referrers joined with Facebook data I showed how to build data-driven personas to inform SEO efforts. I explained our whole method of tagging every keyword with a persona and a need state I wrote the guide building data-driven personas step by step for SEOs EVERY TIME, THE RESPONSE WAS THE SAME: THAT DOESN'T MATTER FOR SEO. 28
29. RAND:“I HATE PERSONAS” TO A PERSONA BUILDER THEN NOW The industry pushed back on the idea. The reality is audience research was never optional. The industry took a decade to understand why it was required. 29
30. WHY CMOS LISTEN CMOs don’t need a better SEO They need a marketer who understands machines. 30
31. I TRIED TO WARN Y’ALL AT SEO WEEK “If we don’t get it together, someone else will own AI Search.” I said this on stage at SEO Week. This isn't hindsight. The people with the right instincts are sitting out the exact moment they should be leading. 31
32. WHAT I LEARNED AT CANNES… I went to talk after talk at Cannes Lions: Adweek, Bain and all the big media players. Every room, every panel, every dinner was about AI Search. Who's winning it. How brands are doing it. Who owns the budget. THERE WERE NO SEOS IN THESE ROOMS None of them were SEOs. The one discipline that's spent 25 years making content legible to machines wasn't at the table where the machine-readable future was being decided. 32
33. I WAS THE ONLY SEO AT THE AI SEARCH TABLE MICHAEL KING · LINKEDIN Just had dinner at Cannes with a bunch of agency owners that do AI Search work. I was the only one from an SEO agency. Everyone else was PR, web dev, creative, digital, or management consultancy, etc. The category is being formed whether the SEO community likes it or not. 223 REACTIONS · 55 COMMENTS · 16,496 IMPRESSIONS 33
34. WHAT THE CMOS AT CANNES THINK OF SEO 34
35. HOW TO TAKE THE LEAD IMPROVING OUR POSITION 35
36. WHAT WE'RE ALREADY GREAT AT ADAPTATION We've rebuilt the playbook through Panda, Penguin, mobile-first, BERT and AI Overviews. Starting over is normal for us. CREATIVITY CURIOSITY With no control over the algorithm and rarely enough budget, we've always engineered our way to results. We take apart patents, leaks and model behavior just to understand how the system works. CROWDSOURCING We test in public and share what we find. The community figures it out faster than any vendor. THESE ARE EXACTLY THE SKILLS AI SEARCH NEEDS. THEY JUST NEED A MARKETING FRAME. 36
37. HOW WE STOP CONFUSING THE MARKET 01 02 03 04 AGREE ON THE JOB AGREE ON MEASUREMENT AGREE ON PROOF PUBLISH THE BUYING CRITERIA Define what the work is and what it isn't, so buyers can compare like with like. Input, channel and performance metrics. Repeated sampling, and the denominator every time. Any visibility claim that isn't based on repeated sampling over time is an anecdote. Give procurement the questions to ask, so the RFP rewards rigor over relabeling. WE HAVE THE POWER OF THE CROWD. NOW IT FOR THE BUYER. Source: iPullRank, "AI Search for the Procurement Professional" (16 RFP questions) 37
38. We need to define what AI Search’s addressable market is. 38
39. A CMO FUNDS MARKETS, NOT PROMPTS PROMPT VOLUME AI demand by topic, from Profound SIZE THE PRIZE Report share of a known market, not share of a hand-picked prompt list. + SEARCH VOLUME Classic search demand by topic PRIORITIZE Put content budget where AI and classic demand overlap. = ADDRESSABLE MARKET Total discovery demand, mapped to entities FORECAST Give finance a number to plan against and a baseline to beat. 39
40. WE NEED TO RECONSIDER PROMPT VOLUMES SEARCH VOLUME PROMPT VOLUMES Search Volume and Prompt Volume data are both modeled from panel data. Why do we consider one usable and the other one not? 40
41. SEARCH VOLUME WAS NEVER ACCURATE EITHER 91% of Keyword Planner volumes were higher than Search Console impressions 54% “There's no such thing as an accurate search volume.” TIM SOULO, CMO, AHREFS Keyword Planner numbers are rounded annual averages, put into buckets, with close variants summed together. We built a whole discipline on a directional number, and it worked. were off by more than half Sources: Ahrefs, “GSC vs. GKP: Comparing Search Volumes for 72k Keywords” (2021); Tim Soulo, “There's No Such Thing as ‘Accurate’ Search Volume,” Ahrefs (2024) 41
42. MOST OF OUR DATA ISN’T “REAL” ANYWAY Almost all SEO data is an approximation that allegedly has some correlation with a metric that Google tracks, which is why companies can randomly change it like this.
43. NO WORKING ADS, NO REAL PROMPT DATA $0.89 vs $2.45 0.37% vs 1.17% 6.6× Revenue per paid visit, ChatGPT vs. Google Search (US) Paid conversion rate, ChatGPT vs. Google Search Top known ChatGPT CPC ($1.72) vs. its value at paid-search parity (~$0.26) GOOGLE'S SEARCH VOLUME EXISTS BECAUSE GOOGLE SELLS KEYWORDS CHATGPT ADS AREN'T PAYING OFF YET SO DON'T WAIT FOR REAL DEMAND DATA. BUILD OUR OWN METRIC. UNTIL THE ADS WORK, WE MEASURE OURSELVES. Source: Kaiser & Schulze, "Can Advertising Fund Free AI?" (working paper, Oct 2026) 43
44. TAM SHOULD BE PRECISE, NOT ACCURATE ACCURATE ON AVERAGE, SCATTERED PRECISE: SAME READING EVERY TIME There is no true number to hit. A TAM model has to rank opportunities and show change over time, and a number that is measured the same way every time does both. Prompt volumes are modeled and directional by design. So is search volume. 44
45. PERSONA-DRIVEN RESEARCH, REBUILT FOR AI SEARCH 2013: THE ORIGINAL PERSONA NEED STATE KEYWORDS CONTENT NOW: SAME PERSONA, MORE SURFACES PERSONA NEED STATE PROMPTS FAN-OUT QUERIES Who they are, from audience research, not assumptions The job they're trying to get done right now How they ask an assistant, in their own words What the engine searches on their behalf ENTITIES CONTENT The concepts both sides share, priced with prompt and search volume Coverage built for the person and the retrieval layer THE PERSONA NEVER CHANGED. THE SURFACES MULTIPLIED. Extends iPullRank's Persona-Driven Keyword Research Process with prompt volumes and query fan-out 45
46. SIZE THE PROMPT PORTFOLIO AT THE ENTITY LEVEL PROMPT VOLUMES FAN-OUT QUERIES SEARCH VOLUME ENTITIES Profound's audience sizing: real, but directional by design The synthetic queries engines run to answer each prompt 95% of fan-out queries have zero monthly search volume One shared unit for AI and classic search demand 46
47. FIVE STEPS TO AN AI SEARCH TAM 01 02 03 04 05 DEFINE THE PORTFOLIO DECOMPOSE INTO ENTITIES FAN OUT EVERY PROMPT ASSEMBLE THE MAP Start with the prompts you already track. Size what you have before you expand it. Break each prompt into entities and entity pairs. Pull prompt volume with phrase match, by platform and intent. Run each prompt across engines and several runs. Extract entities from the synthetic queries the same way. SIZE THEM IN GOOGLE & PROFOUND Pull search volume and prompt volume for the same entities and pairs. Keep the fan-out count as the structural signal. Every entity carries prompt volume, search volume, fan-out centrality and your visibility. SAME EXTRACTION PROMPT ON BOTH SIDES, SO "VOIP PHONE SYSTEM" AND "VOIP" RESOLVE TO ONE NODE Method: iPullRank, "Sizing a Prompt Portfolio" (run as a Profound Agent) 47
48. HERE’S WHAT I BUILT IN PROFOUND I’ve written an agent to that combines search volumes, prompt volumes, and visibility to show what combined addressable market is. Each bubble is one entity from our prompt portfolio. • • • • X-axis: AI prompt volume a month (Profound) Y-axis: Google search volume a month (DataForSEO) Shade: our AI visibility, 0–100 Dashed lines: medians that split the four quadrants Horizontal: AI prompt volume (Profound). Vertical: Google search volume (DataForSEO). 217 entities from iPullRank's own 304-prompt portfolio, Oct 2026. 48
49. WHAT TO DO IN EACH QUADRANT HIGH SEARCH · LOW PROMPT HIGH SEARCH · HIGH PROMPT CLASSIC ONLY DUAL VALUE Demand that never shows up in prompt language. Rank for it the traditional way. Demand in both channels. One piece of content earns classic traffic and AI citations. MOVE: THE CLASSIC SEO PLAYBOOK MOVE: BUILD THESE FIRST LOW SEARCH · LOW PROMPT LOW SEARCH · HIGH PROMPT DORMANT CONVERSATIONAL ONLY No demand in either channel yet. Big bubbles here are central to the portfolio anyway. Real demand with no Google footprint. Keyword tools can't see it, so most teams are blind here. MOVE: WATCH FAN-OUT CENTRALITY MOVE: BUILD FOR RETRIEVAL, MEASURE IN PROMPTS Quadrant lines sit at the medians. Bubble size = number of portfolio prompts touching the entity. 49
50. OUR CATEGORY: “GEO” LIVES IN GOOGLE, NOT PROMPTS CONVERSATIONAL ONLY AI SEARCH STRATEGY CONTENT RELEVANCE RELEVANCE ENGINEERING DUAL VALUE AI SEO CONTENT STRATEGY AI SEARCH ~10M prompts, ~25K searches a month GENERATIVE ENGINE OPTIMIZATION SEO AGENCY DORMANT CLASSIC ONLY Vertical: AI prompt volume (Profound). Horizontal: Google search volume (DataForSEO). Schematic of iPullRank's own 304-prompt portfolio, Oct 2026. 50
51. GET IT IN THE PROFOUND MARKETPLACE Link: https://platform.tryprofound.com/agents/template s?category=tag:Research You need to set up a DataForSEO connector to get the search volume data into the platform. 51
52. BUT WAIT, MIKE… You keep talking about query fan out How can I even get that data in an actionable way? 52
53. QUERYFANOUT.COM MAPS WHAT THE AUDIENCE ASKS FREE Bring your own model keys SIDE BY SIDE OpenAI and Gemini fan-outs compared SHARED INDEX Every run feeds an index the industry can learn from 53
54. FAN OUT QUERIES FOR GEMINI & OPENAI Building off what I started with Qforia, I’ve built a tool that does simulates and performs real searches in the Gemini and ChatGPT APIs to capture query fan outs and give you actionable insights to inform your content engineering efforts. 54
55. IT TELLS YOU WHEN THE MODEL DOES SEARCHES Real searches cost more and sometimes the model doesn’t search. We simulate what the model might generate and compare the two.
56. LET’S RUN A PROMPT Running a prompt pulls top level data on how many synthetic queries are generated, the overlap between models and the cost to run capture the data from the APIs. 56
57. SIDE BY SIDE FAN OUT COMPARISON You get the fan-out queries from both models and where they overlap. The overlapping queries are the ones you prioritize for your content engineering work. 57
58. FAN OUT DIAGRAM You can get a data rich diagram showing the relation of queries, their type and whether the simulations predicted them.
59. GOOGLE QUERY CLASSES All queries are classified into the segments Mark Williams-Cook identified in the SERP data leak. 59
60. PROMPT DRIFT Fan outs are probabilistic, so if we run the prompt multiple times, the tool identifies how things changed.
61. REAL SEARCH QUERIES When the model searches, we capture the exact queries it uses, tells you if they were simulated and allows you to click to perform the search in Google. 61
62. SEMANTIC MAP OF QUERIES We map the data semantically and break it into topics for quick content strategy insights.
63. QFORIA DATA BY QUERY In Qforia we mapped each query to the user’s intent and the model’s content expectation to further inform the content strategy. 63
64. TOKENS & COSTS The tool is “Bring Your Own API,” so you get an indication of token cost per pass with each run you do. 64
65. FAN OUT STABILITY We capture how many queries are generated per run to see how stable the fan outs are per prompt. 65
66. IT’S AN INDEX! All the data captured in the platform is searchable and accessible via an MCP.
67. THERE’S A CHROME EXTENSION TO CAPTURE FAN OUTS CHATGPT GEMINI The Chrome extension captures your fan outs as you use ChatGPT and Gemini and gives you exportable data based on how the interface is actually performing. 67
68. OPENAI+GEMINI FAN OUTS ARE CAPTURED TO THE INDEX THEN NOW If you use the Chrome extension, you can contribute to the dataset by having it sent directly to the index. 68
69. QUERYFANOUT.COM Let’s work together to build a query fan out dataset for everyone 69
70. HOW WE HELP OPERATIONALIZE WHAT BRANDS NEED US TO BE RIGHT NOW 70
71. OUR ROLE Our role right now is more like Marketing management consultant 71
72. SIX ELEMENTS TO DRIVE THIS IMPROVEMENT 01 The Data Layer 02 Emerging Protocols 03 The Content Moat 04 New Metrics 05 Brand Codification 06 Tiger Teams 72
73. ORG REDESIGN: CROSS-FUNCTIONAL TEAMS REQUIRED Most organizations will need to fold or rebuild three or four existing functions to staff this properly. Search, content, data, and brand stop being separate teams optimizing separate KPIs. Do it now, while there's still room to lead. 73
74. THE ONLY QUESTION THAT MATTERS Every discipline is now responsible for the same question: WILL THE AGENT SELECT, CITE, AND SURFACE US ACROSS THE INTERFACES THAT MATTER? IF THE ANSWER'S NO, NOTHING ELSE YOU'RE MEASURING COUNTS. 74
75. Brand ORGANIZE AROUND JOURNEYS, NOT CHANNELS UX Content PRIORITY AI Search CUSTOMER Analytics JOURNEY Technology Product Marketing PR Source footer: How We Reorganized Around AI Search 75
76. GIVE DISCOVERY CLEAR OWNERSHIP EXECUTIVE SPONSOR ACCOUNTABLE PROGRAM clears barriers Source footer: Enterprise AI Search Operating Model LEAD sets priorities CROSS-FUNCTIONAL IMPLEMENTATION WORKING GROUP PARTNERS Plans and delivers build it inside each together business 76
77. ONE MODEL, TWO WAYS TO SCALE MID-MARKET ENTERPRISE OWNER Marketing lead Sponsor + program lead DELIVERY Small working group Central standards + cross-functional brand teams START One buyer journey Selected brands or regions AGENCY PARTNERS: Specialist expertise and delivery capacity TECHNOLOGY: Consistent measurement and repeatable workflows Source footer: How We Reorganized Around AI Search 77
78. MAKE CHANGE MANAGEMENT PART OF THE OPERATING RHYTHM Source footer: How We Reorganized Around AI Search 78
79. WHAT THE PROGRAM NEEDS TO FUND Source footer: How We Reorganized Around AI Search 79
80. WHAT LEADERSHIP CAN ACT ON VISIBILITY CAPABILITY PERFORMANCE EXECUTION Source footer: How We Reorganized Around AI Search 80
81. AI SEARCH IS A BRAND CHANNEL FIRST MEASUREMENT MUST EVOLVE BEYOND PERFORMANCE METRICS Measures how visible, cited, and represented your brand is across search and AI surfaces. • Share of Voice • Named as a source • Accurate about us • Positive about us Signals that show whether your content and site are aligned with how AI and search engines understand and retrieve information. • Quotable content • Known to the machine • Read by the machines • Shows up for buyer questions Outcomes that prove whether visibility and citations translate into business impact. • Traffic • Events • Conversions • Engagement Depth Source footer: How We Reorganized Around AI Search 81
82. WHAT ACTUALLY NEEDS SOLVING RIGHT NOW 01 02 03 04 05 HOW BIG IS THE AUDIENCE? ARE WE EVEN RETRIEVABLE? HOW DO WE MEASURE? WHO OWNS IT? Prompt portfolios are unsized guesses Retrieval decides before any answer is written DO WE HAVE ENOUGH SURFACE AREA? Owned, earned and shared, across every channel Not only as a performance channel. Nobody has decided yet EVERY ONE OF THESE IS A MARKETING PROBLEM. WE NEED TO OPERATE AT HIGHER LEVEL 82
83. WHAT SERIOUS LOOKS LIKE NEXT WEEK 01 SPEAK IN REACH AND BRAND LIFT, NOT RANKINGS AND TRAFFIC 02 BRING ADDRESSABLE AUDIENCE RESEARCH TO EVERY BRIEF 03 OWN THE MEASUREMENT STORY: INPUT, CHANNEL, PERFORMANCE 83
84. DEFINE IT TOGETHER OR GET LEFT BEHIND OPTION 1 OPTION 2 WE COME TOGETHER AND DEFINE THIS THING WE KEEP ARGUING ABOUT THE NAME We set the standards, size the market and lead the cross-functional team. Everything we ever asked for is on the table. Platforms and agencies outside SEO define it for us. CMOs decide SEOs are not serious people, and we get left behind. SO LET'S GET SERIOUS, PEOPLE. 84
85. BE SERIOUS PEOPLE I LOVE YOU. NOW GO BE SERIOUS PEOPLE. serious people. 85
86. ONE MORE THING FOR MY OWN NEED FOR INTELLECTUAL SUPERIORITY 86
87. BUT WAIT, MIKE… You’re always talking about vectors… What do I even do with that data? 87
88. MEET VECTOR WORKSTATION Vector Workstation is our native visual vector analysis tool that ingests Screaming Frog data and does vector analysis. Before AI Search writes a word about you, retrieval has already chosen the passages closest to the query in vector space. VWS embeds your pages, queries and links on your own machine, calibrates the scores so they mean something, and turns every view into ranked actions. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 88
89. SITE FOCUS SCORING WHAT YOU'RE SEEING Each dot is a page, placed by its distance from the site's center. The rings come from our own spread: 422 Core, 839 Focus, 380 Expansion and 40 Peripheral pages. It approximates siteRadius and siteFocusScore from the Google leak. The 37 Peripheral pages no query maps to become prune or refocus actions. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 89
90. TOPIC CLUSTERING WHAT YOU'RE SEEING Each bubble is a topic sized by page count. We have 30, and each has a hub, its most central page. Our largest has 150 pages and 116 clearly covered queries, with the content engineering service page as its hub. The nearest uncovered queries are spam like "rtp pulsa777". Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 90
91. 3D TOPICAL MAP WHAT YOU'RE SEEING Every page in 3D, with "ai mode" (28,000 searches a month) dropped in at its real position. It lands among six of our AI Mode pages scoring 0.733 to 0.755. None clears the 0.76 line, so no page clearly owns the query. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 91
92. SEMANTIC DRIFT WHAT YOU'RE SEEING Every page is checked against the site's center, its own topic and its own title. 88% stay on subject. 46 miss their title, 29 are off-topic and 49 have drifted, and 17 of the off-topic pages are chart HTML files that search engines can crawl. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 92
93. QUERY BULLSEYE WHAT YOU'RE SEEING "ai mode" sits at the center and every page is placed by similarity. Inside the green ring (0.76+) a page clearly targets it; outside the red ring (0.66) nothing answers it. Nothing clears 0.76. Our best page is the AI Mode report at 0.755, with 60 pages in the partial band. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 93
94. VISUAL LINK GRAPH WHAT YOU'RE SEEING The Link Graph focused on our website migration guide, the best page on the site for 23 queries. All five of its editorial links come from copies of one podcast page and are off-topic, while its 22 backlinks score 0.83 for relevance. Other sites understand the page better than we do. VWS names the fix: three related pages that should link to it. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 94
95. VISUAL LINK GRAPH WHAT YOU'RE SEEING The relationship between the internal and external link graph so you can prune for relevance. Source: iPullRank Vector Workstation audit of ipullrank.com, Oct 23, 2026 95
97. FIVE LAYERS, ONE MARKETING STACK 01 AUDIENCE DEMAND 02 MACHINE VIEW 03 QUERY SPACE 04 SURFACE AREA 05 MEASUREMENT Profound prompt volumes, reduced to entities Vector Workstation: retrieval in vector space QueryFanOut.com fan-outs and the shared index Owned, earned and shared media across channels Input, channel and performance metrics HOW BIG IS THE AUDIENCE? ARE WE RETRIEVABLE? WHERE DO WE NEED SURFACE AREA? DO WE HAVE ENOUGH OF IT? HOW DO WE MEASURE IT? 97
98. WRAPPING UP ROLL THE CREDITS 98
99. SCREAMING FROG + OLLAMA I generate embeddings as I crawl, asset content, take screenshots and analyze as I crawl. All on my own local GPU.
100. 25 CHAPTERS OF PURE 🔥🔥🔥 Everything you need to know about how AI Search works. • • • Strategy, Execution, and Measurement Guidance Up to date AI Search statistics Comprehensive Glossary No vagueries.
101. ipullrank.com/ai-search-manual
102. https://relevanceconf.com