This is also using OpenAI's GPT model. So the same hallucinations are probable here for PDFs.