When you perform a fact check with FactSnap, two processes begin simultaneously to analyze your highlighted claim:

1. Preliminary Check: FactSnap sends the text to Groq (a platform that facilitates the interaction with the language model LLama 3.3) and performs a super-fast, short preliminary analysis of the claim. This result is labeled as “Preliminary Check” in your result dialog.

2. Detailed Check: At the same time, FactSnap uses exa.sh to search the web for relevant sources related to the claim. Once exa.sh identifies sufficient resources, FactSnap employs GPT-4o-mini to evaluate the claim against those sources, providing a detailed fact-checked response. This is displayed as “Explanation” in your result dialog.
   
   

After the detailed explanation with supporting sources is ready, FactSnap updates the preliminary result based on the more comprehensive findings.