App or manual? Yes, both. They do different jobs, and the teams that get licensing right use each for what it is good at. - The app is for speed: commercial-use verdicts in seconds, flags known gotchas like Llama's MAU threshold. Triaging ten candidates, that speed is everything. - Manual review is for commitment: once you pick the model you will ship, read the file. Thresholds, training restrictions, redistribution. Twenty minutes, once per version. This is the step that stands up in diligence. - App-only teams ship on summaries and miss version mismatches. Manual-only teams burn hours re-reading Apache 2.0 for the fifteenth time. - Write it down either way: model, version, license, date, source link. A checker that cannot export is a toy. I compare 200 models in plain language, and PrivateLLM deploy sets up the licensed model on your own AWS for $50 plus usage.