Upgrade to Pro — share decks privately, control downloads, hide ads and more …

I Ran 30+ GEO Experiments So You Don't Have To:...

Sponsored · Your Podcast. Everywhere. Effortlessly. Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
Avatar for Thomas Peham Thomas Peham
September 16, 2026

I Ran 30+ GEO Experiments So You Don't Have To: What 200M AI Citations Say Actually Works

Most GEO advice is guesswork. This talk isn't. OtterlyAI ran 30+ controlled GEO experiments and analyzed more than 200M AI citations across ChatGPT, Google AI Overviews, AI Mode, Gemini, Perplexity, Copilot and Claude to find out what really gets brands cited in AI search.

What you'll learn:

Why AI search is a brand channel that also drives revenue, and why GA4 misses most of it (0.1% of conversions in web analytics vs. 10.6% in self-reported attribution)
What happened when we published 2,000 AI-generated blogs: Google deindexed them, but Perplexity, Copilot and Claude kept citing them
Blog vs. YouTube: which format gets cited first, and why 94% of video citations go to long-form videos
Findings from 1.8M YouTube citations, 20M cited URLs (19.3% of them are dead) and a study of AI licensing deals with news publishers
On-page tests: schema, FAQs, markdown, llms.txt, alt tags and URL structure, and which ones help in Google but not in AI search
A practical list of what works in AI search right now and what doesn't

Presented at BrightonSEO San Diego 2026 by Thomas Peham, CEO & Co-Founder of OtterlyAI.

Want our GEO experiment spreadsheet? Send me a LinkedIn DM or email [email protected].

Avatar for Thomas Peham

Thomas Peham

September 16, 2026

More Decks by Thomas Peham

Other Decks in Business

Transcript

  1. Thomas Peham OtterlyAI I Ran 30+ GEO Experiments So You

    Don't Have To: What 200M AI Citations Say Actually Works
  2. BrightonSEO San Diego I Ran 30+ GEO Experiments So You

    Don't Have To: What 200M AI Citations Say Actually Works
  3. Hello. I’m Thomas Peham, CEO & Co-Founder at > ex-VP

    Marketing > in B2B SaaS since 2014 > in Marketing since 2010 Otterly.AI
  4. AI Search The AI Search Funnel Brand Coverage (supported by

    Website Citations) On Page AI Bot/Agent Traffic (Human) Referral Traffic Branded + Direct Traffic (indirect) Conversions Self-reported + web analytics Otterly.AI
  5. Get All GEO Experiments Free GEO Sheet: Send me a

    LinkedIn DM or email ([email protected]) - I’ll send you our GEO Experimentation Spreadsheet Otterly.AI
  6. Form a Hypothesis OtterlyAI Experiment Methodology What GEO tactics &

    myths are we testing, and what result do we expect it to produce? Feedback Community Publish Results Measure Relentlessly Did our experiment move the needle, and what can other GEO practitioners learn from it? Are we tracking AI bot visits, brand mentions, and AI citation rates over a fixed time window? Otterly.AI
  7. OtterlyAI What works in AI Search right now • •

    • • • • • • • • • Press Release Distribution Local + EN Content Duo On-Page Schema Markup for Google AIO Youtube Video Optimization Wikipedia Presence Reddit Brand Visibility on “active” communities (with replies) Linkedin Pulse articles External Listicles Industry Recognition (awards, analysts) Individual “Features” Pages AI-assisted content generation • • • • • • • Clean structured URLs Blog / Editorial Articles for TOFU discovery Product Page & Category Page for Product evaluation & BOFU Documentation, Guides for Product Research Alternative vs pages AI memory Injection AI videos in YouTube What doesn’t • • • • • • • • AI Slop Llms.txt File Markdown Files as variants Schema Markup doesn’t work for ChatGPT, Gemini Image alt tags & file names for AI Search URL structure & length doesn’t matter Beautiful design of content LinkedIn regular feed posts (works less than Pulse) Otterly.AI
  8. Version A (Control) = 1000 Blogs Version B (Treatment) =

    1000 Blogs + Author + Author box Otterly.AI
  9. Four Data Layers 1. Google Data (GSC) 2.Bing Data (BWT)

    3. AI Citations (OtterlyAI) 4.Agent Analytics (OtterlyAI) Otterly.AI
  10. capybaraverse.com Dead after 90 days A Died 3 months later

    1000 AI Blogs with 1 click. 1. Google Data capybara-world.com Dead after 14 days Died 2 weeks later B Otterly.AI
  11. capybaraverse.com A 1. 2. 3. Died 3 months later Google

    Data Bing Data AI Citations capybara-world.com Died 2 weeks later B Otterly.AI
  12. AI Agent Page Visits Over Time A Died 2 weeks

    later Agent Analytics (OtterlyAI) lets you view 4. AI Crawler Bots AI Agent Page Visits Over Time B Otterly.AI
  13. capybaraverse.com A Died 3 months later ChatGPT crawls but citations

    = 0 • AI Citations capybara-world.com Died 2 weeks later B Otterly.AI
  14. • Not every AI Search Engine treated it equally ◦

    Only works on: Perplexity, Copilot, Claude • Bing vs ChatGPT vs Copilot ◦ Bing still indexes our AI-generated content. ◦ Copilot still cites our AI-generated content. ◦ ChatGPT not so much, rarely. • Agent Crawls vs Citations: Just because it gets crawled, doesn’t mean it gets cited • Google's own ecosystem seems to be inconsistent: AI Overviews cite it, Organic fully deindexed AI Slop Experiment Otterly.AI
  15. Video vs. Blog: We Published the Same Story in Both

    Formats. Here’s What AI Search Cited Otterly.AI
  16. Video vs. Blog: We Published the Same Story in Both

    Formats. Here’s What AI Search Cited • Written content gets detected and cited earlier • Every AI Search engine cited the blog version. Not one skipped it. • Text is also what makes the video itself citable • Ship every story twice, article and long-form video, on the same day. • Measure per platform. Perplexity and ChatGPT behave like different universes. • Do not bet on Shorts alone. 94% of video citations go to long form. Otterly.AI
  17. The YouTube Study for AI Search • 100M citations over

    30 days • YouTube 2nd biggest social media channel cited • 1.8M Youtube Citations Otterly.AI
  18. The Results Popularity metrics (likes, views, comments) don’t matter for

    AI Search Active channels have more videos = more chance of being cited Otterly.AI
  19. Case Study Results Share of Voice Increase after 1 Month

    Google AI Mode MS Copilot +114% 14 → 30 +17% Gemini Perplexity +3% 29 → 30 +15% ChatGPT 36 → 42 +83% 40 → 73 AI Overviews 54 → 62 -37% 38 → 24 19 of 25 videos also ranked #1 in regular Google Search. Otterly.AI
  20. We analyzed 20M Cited URLs Dead Citation Study • The

    cited URL and its domain • The AI Search Engine that cited it (ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini, Microsoft Copilot, Claude) • The HTTP status the URL returns when fetched: live, redirect, missing, server error, or unreachable Otterly.AI
  21. Of 20,092,769 cited URLs checked, 80.7% return a live page

    and 19.3% do not • Wikipedia: cited pages that are no longer articles • Of the cited PDFs, 10.66% of the total, are already dead • Reddit: dead citations after the crackdown Otterly.AI
  22. 8 months after pages/URLs were removed. Still 5% citation share.

    • AI builds perceptions from outdated content you or someone else already killed • Your content footprint on AI is bigger than your actual website • Think twice before publishing - AI forgets slower than you think Otterly.AI
  23. Summary of On-Page Experiments Learning Google AI Search Schema Schema

    lifted Google AIO 1,500% but ChatGPT citations dropped 👍 ❌ URL length URL Optimization doesn’t matter for AI Search 👍 ❌ Page Types Guides, Blogs, Help, Home Page as most favorable page types 👍 👍 Images No impact on AI Search Filenames hallucinated 👍 ❌ Alt Tags Only processed by ChatGPT, no other AI Search 👍 partially-👍 Markdown No measurable impact ❌ ❌ FAQ Proven results both on Google + AI search engines 👍 👍 Otterly.AI
  24. Schema lifted Google AIO 1,500% but ChatGPT citations dropped AI

    Search Platform AI Citation Increase or Decrease? ChatGPT decrease AI Overviews increase AI Mode increase Gemini decrease Perplexity No impact Copilot decrease Otterly.AI
  25. Does a deal with a news publisher earn more AI

    citations per page? Otterly.AI
  26. Conclusion: What This Means for PR Teams • News Is

    7.2% of All AI Citations (2nd biggest source on Claude) • Yes. News publishers with OpenAI Deals: +48% More Citations • However, you Do Not Need a Licensed Publisher to Get Cited Analyze the industry first!! • Niche/Trade Media > Expensive Mainstream media • Content Types cited the most: ◦ Top/Best listicles; ◦ How-to-guides; ◦ Problem-discovery; ◦ Product/service reviews Otterly.AI
  27. OtterlyAI What works in AI Search right now • •

    • • • • • • • • Schema Markup for Google AIO Youtube Video Optimization Wikipedia Presence Reddit Brand Visibility on “active” communities (with replies) Linkedin Pulse articles Local + EN Content Duo On-Page Press Release Distribution External Listicles Industry Recognition (awards, analysts) Individual “Features” Pages • • • • • • • Clean structured URLs Blog / Editorial Articles for TOFU discovery Product Page & Category Page for Product evaluation & BOFU Documentation, Guides for Product Research Alternative vs pages AI memory Injection AI videos in YouTube What doesn’t • • • • • • • • AI Slop Llms.txt File Markdown Files as variants Schema Markup doesn’t work for ChatGPT, Gemini Image alt tags & file names for AI Search URL structure & length doesn’t matter Beautiful design of content LinkedIn regular feed posts (works less than Pulse) Otterly.AI
  28. Let’s connect Thanks! You’ve been an otterly amazing audience Thomas

    Peham, CEO & Co-Founder at Free GEO Sheet: Send me a LinkedIn DM or email ([email protected]) - I’ll send you our GEO Experimentation Spreadsheet Otterly.AI