Is your AI's answer grounded — or made up?
Paste a source and a model's response, and see live whether the answer is supported by the source. Calibrated on English and Bahasa Indonesia.
Nothing you paste is stored. No LLM is called.
Pick an example or paste a source + response, then check.
Calibrated on English + Bahasa — AUC 0.97 (EN) / 0.999 (ID).
How it works
A real groundedness model, not a mock
This runs a multilingual natural-language-inference model that scores how well a response is entailed by its source. It's calibrated on an English + Bahasa Indonesia evaluation set — AUC 0.97 on English and 0.999 on Bahasa. It measures groundedness against the source you provide; it is not a general fact-checker. No LLM is called and nothing you paste is stored.
Catch hallucinations across every model, in every language.
The same groundedness signal you just tested — on your own traffic, with data kept in-region.