Generative Engines
Systems that combine a conventional retrieval index with a language model: interpret the question, expand it into sub-queries, retrieve live pages, read them, synthesise an answer, and attach citations.
Analogy Mental model
A senior lawyer with five researchers. You describe the case, the researchers return with twenty relevant files, she reads them in seconds and tells you the conclusion, pointing at the page each sentence came from.
In practice What you actually do
The first competence is knowing what each engine retrieves from. The assistant built into a search engine uses that engine's own index and knowledge graph. The one built on a chat platform leans on a different index and its own crawler. The independent answer engines blend several sources and re-rank them.
Where people get burned Pitfalls
Teams that optimise only for the largest search engine are invisible in assistants that retrieve from elsewhere. If a second index does not contain your site, you do not exist in the answers built on it, and nothing in your main console will tell you.
This entry was written in English from the Chinese original, not machine-translated. Read the Chinese version →