# How AI search picks sources: state of the evidence

German title: Wie KI-Suchen Quellen wählen: Stand der Belege

> Verified are the process description for RAG systems (retrieving documents, generating the answer) and Google's prerequisite for AI Overviews. Not verified is the criteria by which individual sources are selected.

## Facts

- **RAG:** Model + searchable knowledge store (Lewis et al. 2020)
- **E-E-A-T:** per Google not a ranking factor of its own

## What is verified

- RAG: Lewis et al. (2020) describe combining a pretrained language model with a searchable external knowledge store from which a retriever fetches passages.
- Google: The prerequisite is that the page is indexed and snippet-eligible. The documentation names no further selection criteria.
- E-E-A-T: According to Google not a ranking factor of its own; the guidance asks “Who, How, Why”.

## What is not verified

Whether structure, authority or freshness steer the selection, and how strongly, is not publicly documented for Google's AI Overviews, ChatGPT or Perplexity. Measuring through citations is possible but yields only individual observations.

## Sources

- Lewis et al.: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (arXiv 2005.11401v4) (primary source)
- Google Search Central: AI features and your website (primary source)
- Google Search Central: Creating helpful, reliable, people-first content (primary source)

Checked on: 24 October 2026.

## Questions

### Which criteria does Google use for AI Overviews?

Apart from indexing and snippet eligibility, Google names no further criteria.

### Does domain authority count?

That is not verified. Google describes E-E-A-T as an evaluation framework, not a ranking factor.

### What is RAG?

Retrieval-augmented generation: a language model uses retrieved documents to generate its answer (Lewis et al. 2020).

## Related

- [Google AI Overviews: what Google documents](https://munichmedia.com/reports/google-ai-overviews.md)
- [SEO vs. AEO: difference and context](https://munichmedia.com/reports/aeo-vs-seo-unterschied.md)
- [Direct answers for answer engines](https://munichmedia.com/reports/direktantworten-answer-engines.md)
- [Schema.org for AEO: what is verified](https://munichmedia.com/reports/aeo-schema-org-basis.md)

Canonical HTML: https://munichmedia.com/reports/aeo-quellen-ai-overviews