ML Models & Algorithms
Embedding Model
In shortAn embedding model converts text, images or other content into vectors of numbers so that items with similar meaning end up close together.
An embedding model is a machine learning model that turns content into an embedding: a fixed-length list of numbers (a vector) that represents its meaning. Texts, images or products with similar meaning get vectors that sit close together, which lets you compare, cluster and search content by meaning rather than by exact words. Most embedding models are encoder models built on the transformer architecture.
How it works
You pass in a piece of content and the model returns a vector, for example 384 or 768 numbers long. The model has been trained, usually on huge numbers of pairs of related texts, so that related items produce similar vectors. You then measure how close two vectors are, most often with cosine similarity.
Embedding models come in a few families:
- Open-source sentence models, run locally through the Sentence Transformers library. Sentence-BERT popularised the approach; all-MiniLM-L6-v2, the default in BERTopic, produces 384-dimension vectors, and BAAI/bge-small-en-v1.5 is a small, fast model MLforSEO uses on Colab’s free CPU.
- Commercial API models from providers such as Google, OpenAI and Cohere, which you call over an API and pay for per token.
- Multimodal models that place text, images and sometimes video and audio in the same vector space. Vertex AI offers one, and Google launched Gemini Embedding 2 in March 2026 as a fully multimodal model.
How it’s used in SEO and marketing
- Internal linking: embed every page and paragraph, then suggest links between close pairs that are not yet linked.
- Keyword clustering and topic modelling, for example in BERTopic.
- Keyword-to-topic mapping and redirect mapping by meaning when wording differs.
- Related-post modules and on-site semantic search.
- Image SEO: matching images and text, and auditing whether image content matches the page.
- Understanding AI search, where retrieval systems embed chunks of your page to decide whether they match a query.
Example
MLforSEO’s internal linking notebook embeds each page’s title plus its first few hundred words with bge-small-en-v1.5, clusters the vectors with k-means to find topic hubs, and ranks unlinked page pairs by cosine similarity to produce link suggestions for an editor to review.
Limitations and things to watch
- Vectors from different models are not comparable; re-embed everything if you switch models.
- Most models have an input limit, so long pages are truncated or need chunking.
- Similarity scores have no universal threshold. Calibrate on a sample you have checked by hand.
- Choose by task and language; public benchmarks such as MTEB help shortlist models, but test on your own data.
Learn it on MLforSEO
Where Embedding Model shows up on MLforSEO
Read about it on the blog
Sources and further reading
- 10 Ways to Automate Internal Linking with AI MLforSEO
- Topic Modelling for SEO with BERTopic MLforSEO
- Image SEO in 2026: The Complete Guide to Image Optimisation, Advanced Tactics and Multimodal Search MLforSEO
- Sentence Transformers documentation
- Reimers and Gurevych (2019): Sentence-BERT
- MTEB: Massive Text Embedding Benchmark leaderboard
Other terms from ML Models & Algorithms
- BERT (Bidirectional Encoder Representations from Transformers)Google's 2018 transformer language model that reads text in both directions to produce context-aware embedding…
- BERTopicA modular topic modelling technique that clusters transformer embeddings and labels each cluster with class-ba…
- BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies)BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) clusters very large numeric datasets by s…
- BM25A classic ranking function that scores documents by how often query words appear, adjusted for word rarity and…
- Boyer-MooreA fast exact string-search algorithm that compares a pattern from right to left and skips ahead after mismatch…
- c-TF-IDFClass-based TF-IDF: treats each cluster as one document to find the words that best describe it, used by BERTo…
- Centroid-based ClusteringA clustering family that groups data around central points (centroids), each the average of its cluster's memb…
- Convolutional Neural Network (CNN)A deep learning architecture that scans data with small learned filters to detect local patterns; the classic…
- DBSCANDensity-based clustering that groups closely packed points, finds the number of clusters itself and labels iso…