Guardrails
The limits placed on what an AI may say or do — refusing certain requests, staying on topic, never inventing a price. Essential on anything customer-facing.
Related terms
Query Fan-Out
When an assistant turns one question into several searches behind the scenes, then combines what it finds. It means your page can be pulled into an answer about a question you never targeted.
Human-in-the-Loop
Keeping a person in the decision, either approving what the AI produced or handling the cases it flags. The usual sensible middle ground between doing everything by hand and trusting a machine completely.
Semantic Search
Search that works on meaning rather than exact words, so a page can match a query it never literally contains. It is why writing naturally now beats repeating a keyword.
Hallucination
When an AI states something false with complete confidence — an invented price, a policy you never had, a product you do not sell. The usual cause is that the correct fact was not available anywhere it could read.
Content Negotiation
When the same URL serves different formats depending on what the visitor asks for — HTML for a browser, Markdown for an assistant. It lets one address serve people and machines without either getting a compromise.
Answer Engine
An answer engine replies to a question with a written answer instead of a list of links. ChatGPT, Perplexity and Google’s AI Overviews are all answer engines. The old question was “do we rank?”; the new one is “are we the answer?”