Known vs Found: does AI know your brand?
Known vs Found is in development: a check that separates what AI models know about a brand from training from what they find when they search.
In development: not available yet
What Known vs Found means
An AI engine can put a brand in an answer in two ways. It can already know the brand from its training data (parametric knowledge), or it can find the brand by searching the web while it answers (retrieval). The two have different causes and different fixes, so they are worth measuring apart.
Trained knowledge changes only when a new model ships, and it grows with how often a brand appears next to its category in text across the web. Search results change weekly and depend on crawlable, indexed pages that answer the searches the engine runs. A brand that is known but not found needs better pages; a brand that is found but not known needs more off-site mentions, which take months to reach a model.
Status: in development
The Known vs Found check is being built and is not available in optmz yet. Today, optmz tracks answers with web search turned on, which mixes both sources, and shows the web searches each engine ran so you can see what it looked for. This page will be updated when the check ships.
How the check is designed to work
- Ask the same buyer questions with search off (what the model knows) and with search on (what it finds), and report both rates with their ranges.
- Ask models factual questions about the brand, such as what it sells and who it competes with, and flag answers that are clearly wrong.
- Show which model version each number comes from, because trained knowledge is frozen per release.
The research behind it is in the guide why your brand isn't in ChatGPT answers.