What is tool use?
Tool use is a model's ability to call an outside function mid-answer, such as a search, a calculator, or a code runner, and fold whatever it gets back into its response.
On its own, a language model only predicts text: it can't actually browse a live page, run exact arithmetic, or query a database. Tool use is what a product adds on top: a defined set of functions the model is allowed to call, along with the judgment to decide when a question needs one. The model calls the tool, reads what comes back, and continues its answer with that result in hand.
This is the mechanism behind most grounded, source-backed AI answers: when an assistant cites something, it usually means a tool ran during that exchange, most often a web search, and returned a page for the model to read, rather than the fact being memorized from training. Which tools a model is even allowed to reach for on a given question is set by the product's system prompt, not decided freely by the model or the user's phrasing alone.
Related
- Web search toolThe web search tool is the specific function that lets an AI assistant fetch and read live web pages mid-conversation, and it's what actually produces a citation rather than a memorized guess.
- GroundingGrounding is when a model answers using documents retrieved at the time of the question, rather than only from what it absorbed during training.
- System promptA system prompt is a set of instructions a product's developers attach behind every conversation, invisible to the user, that shapes how the model behaves before it ever sees a question.
- Retrieval-augmented generation (RAG)Retrieval-augmented generation is the technique of retrieving relevant documents and giving them to a model as context before it answers.
Want to know where you actually stand on this? Run a free visibility check or try the free tools.