Unlocking the hidden potential of vector embeddings in international SEO

In this presentation Frank explains how you can use vector embeddings in international SEO. He explains the technique behind it and how you can use it for mapping hreflang tags, internal linking and content optimizations.

Transcript

2. Connect with me on Frank van Dijk Background story about me I spent the summer underwater
3. Connect with me on Frank van Dijk Guiding the dives Showing cool stuff underwater Communicate with other divers Becoming a divemaster means
4. Connect with me on Frank van Dijk Divers speak another language We use hands to communicate 0.9 0.7 0.2 0.6
5. Connect with me on Frank van Dijk What are embeddings? “Embeddings are numerical representations of data (like words, images, or audio) in a multi-dimensional space” Images Audio Text Embedding model 0.9 0.7 0.2 0.6
6. Connect with me on Frank van Dijk cat Embedding model 0.9 0.7 -0.3 0.6 Loyalty Dangerous C uddliness Living being
7. Connect with me on Frank van Dijk cat dog 0.9 0.7 -0.3 0.6 0.9 0.6 -0.2 0.8 Cosine similarity
8. Connect with me on Frank van Dijk cat pet lion dog
9. Connect with me on Frank van Dijk Text-embedding-3-large By OpenAI, highest performance but more expensive Text-embedding-3-small By OpenAI, excellent performance and lower cost Gemini-embedding-001 By Google, flexible in use with dimensions Comparison of different models
10. Connect with me on Frank van Dijk Hello Hola Bonjour Hallo Ciao こんにちは What about international SEO
11. Connect with me on Frank van Dijk Hello Hola Bonjour Hallo Ciao こんにちは 0.9 0.7 -0.3 0.6 Embeddings holds the semantic value
12. Connect with me on Frank van Dijk No, it doesn’t make you a native speaker. But yes, it can help you overcome a language barrier
13. Connect with me on Frank van Dijk Big websites ask for creativity Comparing 1k+ urls in a content gap… No way 🫠
14. Connect with me on Frank van Dijk I mean... It shows you interesting keywords, but not the whole picture Content gap 🫤
15. Connect with me on Frank van Dijk Try Plotly for your visualisations Free Python package Interactive data showcase Web app integrated
16. Connect with me on Frank van Dijk
17. Connect with me on Frank van Dijk Organic traffic data Vector embeddings Competitor & own urls
18. Connect with me on Frank van Dijk API Access >> Ahrefs + Gemini API connection
19. Connect with me on Frank van Dijk Ahrefs >> Backlinks, RefDomains, URL Rating off and Traffic on Gemini >> Extract embeddings from page content Right settings
20. Connect with me on Frank van Dijk Extraction >> Store rendered HTML Rendering >> JavaScript Rendering on
21. Connect with me on Frank van Dijk Export to graph with Python
22. Connect with me on Frank van Dijk Content gap Ranking gap
23. Connect with me on Frank van Dijk Perfect showcase of the right data Show your strengths Show optimization options Makes it visual for non SEO
24. Connect with me on Frank van Dijk What about internal links?
25. Connect with me on Frank van Dijk Enable embedding functionality, semantic similarity and low relevance Connect to your embedding template With embeddings
26. Connect with me on Frank van Dijk 0 1 Identical No similarity Cosine similarity Interesting for a link 0.8
27. Connect with me on Frank van Dijk Try a threshold of 0.8 Crawl to find new internal links
28. Connect with me on Frank van Dijk Similar pages will show up Source url Target url
29. Connect with me on Frank van Dijk Workflow to find non existing links Inlinks Semantically similar
30. Connect with me on Frank van Dijk Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Audit anchors with cosine similarity 0.9 0.7 -0.3 0.6 0.9 0.6 -0.2 0.8
31. Connect with me on Frank van Dijk Finding irrelevant anchors Inlinks Embeddings
32. Connect with me on Frank van Dijk Look further then quick wins Relevant content with link Page Page Page Try and challenge developer to make smart templates This generates real value for users
33. Connect with me on Frank van Dijk Hreflang mapping can be difficult Changing URL per language No 1 to 1 text match Big websites make it harder
34. Connect with me on Frank van Dijk Pages with the same content will have the same semantic value Hello Hola Bonjour Hallo Ciao こんにちは 0.9 0.7 -0.3 0.6
35. Connect with me on Frank van Dijk 0.9 0.7 -0.3 0.6 Fiets Bicicleta Bike 0.9 0.7 -0.3 0.6 0.9 0.7 -0.3 0.6
36. Connect with me on Frank van Dijk 0.9 0.7 -0.3 0.6 Fiets Bicicleta Bike 0.9 0.7 -0.3 0.6 0.9 0.7 -0.3 0.6
37. Connect with me on Frank van Dijk Mapping hreflang tags Embeddings
38. Connect with me on Frank van Dijk Vector embeddings are so interesting because… Google has been using them for years
39. Connect with me on Frank van Dijk Practical examples Chatbots Recommendation systems Search engines
40. Connect with me on Frank van Dijk Retrieval Augmented Generation Identify Retrieve information Generate AI Searching for information most similar to the query
41. Connect with me on Frank van Dijk We need to check what content is used to generate the AI overview +more sources
42. Connect with me on Frank van Dijk How good is their content based on a semantic level +more sources 0.78 0.82 0.81
43. Connect with me on Frank van Dijk Should we optimize for a cosine similarity of 1.0?
44. Connect with me on Frank van Dijk Don’t forget your SEO It’s the foundation of a successful GEO strategy. Without a strong SEO base, your GEO strategy will fail
45. Connect with me on Frank van Dijk Success is no longer just about matching a query, it’s about deeply understanding the intent and the questions behind it, and answering them
46. Connect with me on Frank van Dijk PHASE 1 PHASE 2 PHASE 3 PHASE 4 PHASE 5 Find relevant queries Extracting content Generate embeddings Calculating (cosine) similarity Optimizing content
47. Connect with me on Frank van Dijk Start by finding the full search intent Get insights from:
48. Connect with me on Frank van Dijk We can take a look behind the scenes at the competition with Ahrefs
49. Connect with me on Frank van Dijk Use regex to filter out long tail keywords that might be relevant to your query
50. Connect with me on Frank van Dijk Or check what ChatGPT is using to find information to generate an answer
51. Connect with me on Frank van Dijk All these insights give us a better insight of what our target audience is looking for SERP features like people also ask, people also search for Insights from Gemini Trends from Google Trends
52. Connect with me on Frank van Dijk Extract the right content to compare Get the content by using:
53. Connect with me on Frank van Dijk We need to scrape the content that is used in the AI overviews
54. Connect with me on Frank van Dijk Chunk your own content so we can use it check if it’s good (enough)
55. Connect with me on Frank van Dijk Calculate the cosine similarity 0.9 0.7 -0.3 0.6 0.9 0.6 -0.2 0.8
56. Connect with me on Frank van Dijk One big workflow
57. Connect with me on Frank van Dijk Please don’t start keyword stuffing now…
58. Connect with me on Frank van Dijk Start adding value instead
59. Connect with me on Frank van Dijk What’s next Start calculating cosine similarity Switch from search query to intent Start automating with the Google Sheets 01. 02. 03.
60. Connect with me on Frank van Dijk Your [SEO/GEO/AEO] strategy should still draw the map and set the destination, let vector embeddings be the compass that guides you throughout the journey