# Vector index for your own content: basics

German title: Vektorindex für eigene Inhalte: Grundlagen

> Whoever wants to make own content searchable for an own assistant can split it into passages, add metadata (title, type, date, ID), convert it to vectors and store it in a vector index. Lewis et al. (2020) describe this principle with a Wikipedia index. It is not a method to influence public answer systems.

## Facts

- **Principle:** Retrieval of similar passages from a vector index
- **Scope:** own assistants, not public systems

## Procedure (technical description)

- Split articles into passages, for example at headings.
- Add metadata to each passage: page URL, heading, date, entity ID.
- Convert passages to vectors (embedding model) and store them in a vector index.
- On a query, retrieve similar passages, optionally filter by metadata, and generate the answer with source references.

Chunk size and embedding model are project decisions; changing the model changes the similarity values.

## Scope

This technique concerns your own search functions and assistants. That it influences public answer systems is not verified.

## Sources

- Lewis et al.: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (arXiv 2005.11401v4) (primary source)

Checked on: 24 October 2026.

## Questions

### Does this replace classic SEO?

No. It concerns your own search functions.

### Which chunk size is right?

There is no generally verified value; it must be tested per content and model.

### Do entity tags improve retrieval?

This can be tested for your own index (benchmark on your own dataset). No general evidence is available.

## Related

- [Entities and information gain: concept](https://munichmedia.com/reports/aeo-entity-information-gain.md)
- [Schema.org for AEO: what is verified](https://munichmedia.com/reports/aeo-schema-org-basis.md)
- [Brand retrieval & RAG: what can be influenced](https://munichmedia.com/reports/aeo-brand-retrieval-rag.md)
- [Headless CMS: quality checks in the build](https://munichmedia.com/reports/aeo-headless-cms-pipeline.md)

Canonical HTML: https://munichmedia.com/reports/aeo-vektor-entity-graph