Glossary
The words that keep appearing when people talk about being cited by AI search.
- GEO (generative engine optimization)
- The practice of structuring content so an AI answer can retrieve a passage, trust it and quote it with attribution.
- AEO (answer engine optimization)
- Overlaps with GEO; usually used for the same work on shorter factual answers and voice assistants.
- AI Overview / AI summary
- A generated answer shown above classic results. The engine writes it from retrieved passages and links a few sources.
- RAG (retrieval-augmented generation)
- The pattern behind most AI search: fetch passages first, then generate the answer from them instead of from the model alone.
- Citation
- A link from the generated answer to the source it used. It is the unit GEO is optimising for.
- Grounding
- Tying each claim in an answer to retrieved text. Ungrounded answers hallucinate; grounded ones cite.
- Chunk / passage
- The slice of a page that retrieval indexes and scores. Chunk boundaries often ignore your paragraphs.
- Embedding
- A vector representation of a passage, used to match it to a question semantically rather than by keyword.
- Crawler (AI)
- A bot that fetches pages for training or for retrieval. Its user agent appears in your access log; a 403 there is a total block.
- llms.txt
- A plain-text file at the site root listing the pages worth reading. A convenience signal, not a ranking factor.
- Prompt set
- The fixed list of real user questions you re-run over time to see whether answers cite you.
- Entity
- A named thing — person, product, place — that engines can associate with a topic. Clear entity signals make attribution easier.