ANALISIS EFEKTIVITAS DAN AKURASI CHATGPT-5.5 DALAM PROSES TAKHRIJ HADIS: STUDI KOMPARATIF ANTARA METODE TAKHRIJ MANUAL DI PRODI ILMU HADIS FAKULTAS USHULUDDIN DAN STUDI ISLAM UIN SUMATERA UTARA

Abstract

This study aims to evaluate the performance of ChatGPT-5.5 in conducting takhrij hadith through a comparative analysis with manual methods based on al-Mu‟jam al-Mufahras li-Alfāẓ al-Ḥadīth an-Nabawī referencing classical primary books. The assessment focuses on accuracy (source conformity) and effectiveness (clarity of information and completeness of data variation), rather than testing hadith quality. The results indicate that ChatGPT-5.5 operates at a level of High-Efficiency with Low-Reliability. In terms of effectiveness, this AI excels in presenting information instantly, systematically, and communicatively. However, regarding accuracy, ChatGPT-5.5 suffers from structural limitations leading to massive micro-level index hallucination, characterized by widespread errors in hadith numbering across Sunan books, distortion of narrators‟ identities, fictitious anachronisms, and text transcription errors of the Qur‟an. This occurs because the LLM architecture relies on statistical word probability rather than rigid numerical data. The final conclusion finds that ChatGPT-5.5 cannot be used as a sole reference; it must be positioned as a heuristic tool that requires complementation through manual verification methods.