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
Unsupervised Learning
Training a model on unlabelled data and letting it find structure by itself — grouping similar customers, say, without being told what the groups are.
llms.txt
An emerging convention: a plain-text file at your site root that gives AI systems a clean map of what is on it. It saves an assistant from crawling around to work out what your site contains.
Sentiment Analysis
Automatically judging whether text is positive, negative or neutral. Used to read reviews and messages at a scale nobody could read by hand.
Author Authority
The weight given to content because of who wrote it. Advice attributed to a named person with a real background is treated very differently from advice signed by a company or by nobody.
Generative AI
AI that produces something new — text, images, code, audio — rather than only classifying or predicting. The generative part is why it can write an answer about your business instead of just linking to it.
Text-to-Speech
Turning written text into spoken audio. Modern voices are convincing enough to be used for real narration, which raises its own questions about disclosure.