Retrieval-augmented generation (RAG)
Retrieval-augmented generation (RAG) is a technique in which a system retrieves relevant documents at query time and has a language model generate an answer grounded in them. It was introduced in a 2020 paper by Lewis et al. Google says its AI features use RAG, “also known as grounding,” over its Search index. For visibility, it means your page must be retrieved before it can be cited.
Where does RAG come from?
“Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks” by Patrick Lewis and colleagues was first posted on May 22, 2020.[1] It combined a retriever with a generator so answers could draw on an external document index.
How do AI search engines use RAG?
Google defines RAG as relying on core Search ranking systems “to retrieve relevant, up-to-date web pages from our Search index,” then generating a response with links.[2] The GEO paper models generative engines the same way: retrieve sources, then generate a cited response.[3]
Why does RAG matter for search visibility?
- No retrieval, no citation. Pages must be crawlable and indexed by the engine’s system; see AI crawlers.
- Sub-queries matter. Engines may expand questions; see query fan-out.
- Passages matter. The model needs a passage that supports its claim; see AI citations.
Does RAG stop hallucinations?
It reduces but does not eliminate unsupported claims: in a 2023 audit, about half of generated sentences were fully supported by their citations.[4] See AI hallucinations about brands.
Frequently asked questions
What is grounding in AI?
Generating an answer from retrieved documents. Google uses “grounding” as another name for RAG.
Is RAG the same as training?
No. RAG uses documents at answer time; training changes the model’s weights.
Do all AI chatbots use RAG?
Those with web search do when they search; answers without search rely on training data.
See also
References
Sources accessed October 14, 2026 unless another date is given. See all sources.
- “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al.)”. arXiv. v1 May 22, 2020.
- “Optimizing your website for generative AI features on Google Search”. Google Search Central. accessed October 14, 2026.
- “GEO: Generative Engine Optimization”. arXiv (Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan, Deshpande). v1 submitted November 16, 2023; published at KDD ’24.
- “Evaluating Verifiability in Generative Search Engines”. arXiv (Liu, Zhang, Liang; Stanford). April 2023; Findings of EMNLP 2023.