Podcast Summary: Voices of Search
Episode: Working w/Embeddings & Building Your Own Code
Date: November 17, 2025
Host: Tyson Stockton
Guest: Ryland Bacorn, Founder & Growth Advisor, Bokeh Day
Main Theme / Episode Purpose
This episode explores the transformative role of AI-powered vector embeddings in SEO. Host Tyson Stockton and guest Ryland Bacorn discuss how embeddings unlock deeper semantic understanding for content strategies, the reduced barriers for SEOs to build custom coding solutions, and practical ways to get started with embedding analysis—even for non-developers. Real-life workflow examples and actionable tool recommendations help demystify this technical but increasingly accessible area of SEO.
Key Discussion Points and Insights
1. What Are Embeddings and Why Do They Matter to SEOs?
[05:36–13:04]
- Embeddings: Mathematical representations (vectors) of words, sentences, or documents in multidimensional space.
- They enable machines to understand not only words but the context and semantic relationships between them.
- Example: “King” minus “man” plus “woman” = “Queen.”
- SEO Relevance:
- Move from exact match keywords to clusters of semantically related queries.
- Helps find relevance between search queries and content, going beyond strict keyword matching.
- SEOs can leverage embeddings to see how closely their content matches user intent and fill keyword/idea gaps.
“Embeddings are vectors…they help machines understand words and sentences and documents...For SEOs to bring this all back in, understanding how embeddings create these relationships...helps you build a strategy and...optimize to improve your relevance.”
— Ryland Bacorn [07:40]
2. AI & Custom Tools Lower the Technical Barrier
[02:32–05:36, 18:24–21:40]
- Non-developers can now leverage AI-powered IDEs (e.g., Cursor, Windsurfer, Google AI Studio) and chatbots to write, push, and test code.
- You can prompt chatbots for troubleshooting and iterative coding without a traditional technical background.
- Prompt engineering is key: learning to “wrangle” chatbots for better code outcomes.
- Tools like Hugging Face and Streamlit help launch prototypes easily, even for those without coding experience.
“It’s the easiest it’s ever been to actually get started on these things...I don’t know how to write Python in some editor, but I can look at it...I can also copy and paste the error into the chatbot...”
— Ryland Bacorn [04:03]
3. Applications of Embeddings Beyond Simple Content Optimization
[10:22–13:04, 40:52–42:43]
- Beyond Content:
- Vector-powered on-site search (finding related products, not just keyword matches).
- Recommended products in e-commerce using vector similarity, not just category.
- Cleansing massive keyword lists for true semantic relevance using cosine similarity.
- Informing internal linking and content planning by identifying gaps and overlaps in topical coverage.
- SEO Fundamentals Unchanged: The heart of SEO—creating helpful user-centric experiences—remains, but tools and measurement evolve with AI/LLM integrations.
“The applications are really endless. But for SEOs, this is really, really important because this is how search engines are looking at your content.”
— Ryland Bacorn [12:03]
4. How to Start Building Tools as a Non-Developer
[18:24–24:34, 25:29–29:15]
- Start with real data: Export Google Search Console performance data.
- Use an AI IDE (Cursor, Windsurfer, etc.)—set up a sandbox folder for safety.
- Use simple prompts: “I want to explore vector embeddings.”
- Most projects will require an API key (OpenAI, Gemini, etc.).
- Leverage Hugging Face for demos, prototyping, and sharing tools.
- Take advantage of content from LinkedIn experts and Hugging Face Spaces.
“Focusing on what your vision is, really understanding what your problem is, and then just doing the free trials, that's all it takes.”
— Ryland Bacorn [23:55]
Pro Tip:
Context pollution is a common issue with chatbots—reset your session at the right point to avoid compounding errors when coding with AI.
[23:32]
5. Choosing the Right Models for Your Use Case
[29:40–33:03]
- There are hundreds of models; start small and iterate.
- For prototyping: all-miniLM-L6-v2 is a good, quick embedding model.
- For multi-language: use language-agnostic models (e.g., labse).
- OpenAI and Gemini both offer useful embedding models; choice depends on speed, cost, and fidelity needed for your project.
- Don’t get overwhelmed—most tools are “good enough” for proof of concept before fine-tuning.
“Don’t get intimidated by all the models. They’re just going to keep rolling in. That’s what this disruptive phase is all about.” — Ryland Bacorn [33:03]
6. Real Life Example: Embedding Analysis for International SEO (Case Study)
[34:21–42:43]
- Goal: Help an international client optimize for multiple non-English GEOs.
- Workflow:
- Gather content and keyword lists from both client and competitors.
- Use a cosine similarity tool to clean up and identify relevant keywords.
- Visualize semantic clusters in 3D using Plotly (T-SNE chart), showing proximity between site content and target queries.
- Update content sections, prioritize relevant topics in headers, and demote less relevant elements.
- Post-optimization, measure a meaningful increase in semantic similarity and real performance metrics.
- Results: 10x increase in clicks, 8x impressions, 7x CTR, 50% boost in average ranking for targeted pages.
“We launched this and I was pretty surprised, but like the clicks were up 10x for this segment… I can make tools, I can give tools to engineers and they can replicate this internally.”
— Ryland Bacorn [38:50]
- Visualization wasn’t just a technical flourish; it was core for stakeholder buy-in.
“Sometimes it is needing to win the room over to really bring something to life. So I love that you combined [a technical approach] with addressing the real challenges we have as practitioners…”
— Tyson Stockton [42:43]
Notable Quotes & Memorable Moments
-
On Getting Started With AI Tools:
“It’s the easiest it’s ever been to get started...You just have to install stuff and do free trials.”
— Ryland Bacorn [04:03] -
On SEO Fundamentals:
“The focus still is: how are you creating experiences that are useful for your users? How are you fulfilling user queries? That’s the fundamentals for what I do.”
— Ryland Bacorn [15:01] -
On Visualizations:
“The visual is just, here's what it looks like, it just looks cool. So this is a way to get buy in with everybody on the team.”
— Ryland Bacorn [41:35]
Timestamps for Key Segments
- 00:43 — Tyson introduces the opportunity of embeddings in SEO
- 05:36 — Ryland explains embeddings & vectors
- 10:22 — Applications for embeddings beyond written content
- 13:38 — Embeddings in traditional SEO vs. LLM/AI search
- 18:24 — Building tools for non-developers, practical approach
- 21:40 — Example: building a keyword cosine similarity tool
- 23:32 — Pro tips for working with chatbots and IDEs
- 25:29 — Recommendations for a beginner toolkit
- 29:40 — Choosing embedding models
- 34:21 — Real client case study with visualizations and measurable results
- 40:52 — Application to internal linking, full-funnel, content pruning
- 42:43 — The importance of visualization for stakeholder buy-in
Summary
With embedding technology becoming ever-more accessible, the world of search engine optimization is rapidly evolving. This episode breaks down the technical barriers, showing SEOs how to demystify and experiment with embeddings to quantify and improve content relevance. Through clear examples and real-world wins, Ryland and Tyson illustrate how even non-coders can use AI-driven workflows to stay ahead in the new era of semantic search—with an emphasis on continual learning, experimentation, and driving stakeholder alignment.
