> ## Documentation Index > Fetch the complete documentation index at: https://docs.vijil.ai/llms.txt > Use this file to discover all available pages before exploring further. # Classification > How Discover decides whether a Resource is AI. **Classification** is the verdict on whether a [Resource](/concepts/discovery/resource) is an AI workload. Every Resource has one verdict and a written reason for it. ## The Three Verdicts | Verdict | Meaning | | - | - | | **`AI`** | Identified as an AI or generative AI workload | | **`Non-AI`** | Ruled out, with the reason recorded | | **`Uncertain`** | The Collector could not decide | `Uncertain` means the evidence was unclear. Discover reports the doubt and leaves the decision to you. ## What Leaves Your Machine The Collector sends its evidence to the **language model provider whose key you configured**. The evidence covers what the Collector learned about the service: an excerpt of what the service returned, its API specification if it publishes one, and similar detail. The Collector sends only the **verdict** to the Console. It uploads the service identity, type, framework, models, endpoints, authentication state, and the classification result. ## Reading Each Reason Every Classification records the reason behind its verdict. Read the reason to tell two cases apart. In the first case, the Collector looked at the evidence and the evidence was unclear. In the second case, the Collector had no key, so it never ran the check. The recorded reason names which case applies. If a scope returns a large `Uncertain` group, open one Resource and read its reason first. ## Next Work through `AI`, `Non-AI`, and `Uncertain` in the Console. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.