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

Your Competitors Are Prompting. You Could be Bu...

Sponsored · SiteGround - Reliable hosting with speed, security, and support you can count on.

Your Competitors Are Prompting. You Could be Building.

Presentation given at SEO Week 2026 by Sam Torres all about adding machine learning models into your LLM workflows and automations for improved accuracy, analysis, and repeatability.

Avatar for Sam Torres

Sam Torres

May 01, 2026

More Decks by Sam Torres

Other Decks in Marketing & SEO

Transcript

  1. The state of AI in SEO. 01 AI is everywhere

    in SEO. 02 Most work = one-shot prompts. 03 Prompts don't connect to your stack. PR OMPT draft a strategic point of view → OUT PUT one-time project ✕ D EA D E ND not scalable, no system Sam Torres | The SEO Mermaid
  2. Your tools give you data. Acting on it at scale

    is where most teams stall. Sam Torres | The SEO Mermaid
  3. The difference is ML. Vibe Coding / Prompt Engineering ◈

    ◈ ◈ ◈ ◈ Machine Learning ◈ ◈ ◈ ◈ ◈ Sam Torres | The SEO Mermaid
  4. What you’ll leave with 0 1 One framework 0 2

    Two workflows 0 3 Hands-on notebooks Sam Torres | The SEO Mermaid
  5. LLMs. llm _out p ut .j son l → CLUSTER_01

    · WATER SPORTS Competitor owns 3x depth CLUSTER_02 · F OOTWEAR Missing entirely CLUSTER_03 · VENUES Parity reached Sam Torres | The SEO Mermaid
  6. Each layer answers a different question. LAY E R 01

    Embeddings → LAY E R 02 Clustering → LAY E R 03 LLMs Tool-agnostic. Workflow-agnostic. That's the point. Sam Torres | The SEO Mermaid
  7. Alone, each layer breaks. 0 1 Embeddings alone 0 2

    Clustering alone 0 3 LLMs alone CHAIN THEM TOGETHER → AUTOMATION Sam Torres | The SEO Mermaid
  8. Direct & delegate like you have an intern. A bright

    intern who's never seen your stack, your brand, or your stakeholders. Capable, but new. The clearer your brief, the better the work. Sam Torres | The SEO Mermaid
  9. Think 80/20 for effort. Whichever variant works for you, run

    with it: ◈ 20% of the work delivers 80% of the result ◈ Ship the 80% Sam Torres | The SEO Mermaid
  10. Traditional comp gap tells you what. Not how deep. WHAT

    IT TELLS YOU Which keywords you're missing. A list, sorted by volume. WHAT IT DOESN'T How deeply a competitor owns a topic. Where the structural holes are in your content. Sam Torres | The SEO Mermaid
  11. THE BETTER BUSINESS QUESTION NOT THIS THIS INSTEAD "Who owns

    this topic space, and where are the gaps in breadth and depth?" Sam Torres | The SEO Mermaid
  12. Turn pages into math. from sentence_transformers import SentenceTransformer model =

    SentenceTransformer('all-MiniLM-L6-v2') pages = pd.read_csv('crawl.csv') vectors = model.encode(pages['content']) # → (4,218 pages, 384 dims) Sam Torres | The SEO Mermaid
  13. Cluster topics and pages into neighborhood. OUT PUT · C

    LUST E RS .J SO NL water sports 18 / 52 destination guides 6 / 34 gear reviews 0 / 28 venue roundups 22 / 19 you / competitor · pages per cluster Sam Torres | The SEO Mermaid
  14. Breadth × Depth. BREADTH How many clusters you cover vs.

    them. DEPTH Within shared clusters, how much content they have vs. you. Sam Torres | The SEO Mermaid
  15. Internal linking is high-leverage. Most teams still do it by

    hand. Manual Keyword-match tools Result Sam Torres | The SEO Mermaid
  16. Linking sits at the intersection of 4+ functions. CMS Content

    strategy Crawl data Engineering → One ML pipeline Sam Torres | The SEO Mermaid
  17. Same three layers. Different direction of analysis. WO RK FL

    OW 01 Look outward. WO RK FLOW 02 Look inward. Sam Torres | The SEO Mermaid
  18. All about your own site. R EQ UIR E D

    Your crawl export OPT I ONAL Existing link map Sam Torres | The SEO Mermaid
  19. Your content & link graphs, in vectors. The clustering surfaces

    semantic relationships that anchor-text matching would miss entirely — pages that belong together even when they share no keywords. i n ter nal _ grap h .v i z 3 clusters · 2 cross-cluster bridge opportunities detected Sam Torres | The SEO Mermaid
  20. TH E S AME F RA ME WO RK ,

    S I D E B Y S I DE Two workflows. One pipeline. WORKFLOW 01 · COMP GAP WORKFLOW 02 · INTER NAL LINKING INPUT Competitor crawl + GSC Your own crawl QUESTION Who owns the topic space? Where are my linking gaps? EMBEDDINGS Content similarity across sites Semantic similarity within site CLUSTERING Topic neighborhoods: comp vs. you Topical neighborhoods: your content city LLM OUTPUT Gap report + content priorities Linking recs + anchor text Sam Torres | The SEO Mermaid
  21. Some models to try. Data type Business question Free /

    Local Premium Take note Video What topics does this cover? Whisper ST Embeddings AssemblyAI OpenAI Transcript quality, accents & jargon Image What’s in the image? BLIP/CLIP GPT-4o Vision Scale cost PDF Extract & cluster content Tesseract ST AWS Textract OpenAI Scanned vs. text-native Multilingual Cross-language topic analysis Multilingual-miniLM text-embedding-3- small Start with language detection Sam Torres | The SEO Mermaid
  22. Try it with your own data. 01 Competitor gap analysis

    02 Internal linking T HE SEOM ER MA ID > COM/SE O -W EE K-2 0 2 6 Sam Torres | The SEO Mermaid