AI Preaching: Between Technological Convenience and Scholarly Authority

In recent weeks, Indonesian social media has been abuzz with the appearance of an unusual preaching account. A hijab-wearing figure named Nia Hajar—with a gentle voice, symmetrical face, and distinctive dimples—delivered Qur’anic verses, hadith, and Islamic advice to nearly 900,000 followers, garnering millions of views. It later emerged that the figure was not human; she was a persona entirely constructed using artificial intelligence.

This phenomenon does not merely demonstrate technological progress. It poses a far more fundamental question: when AI begins to speak on behalf of religion, to whom should the community entrust its knowledge?

AI indeed offers many tangible benefits. The technology can translate lectures into various languages, generate automatic subtitles, summarise classical texts, and help people with disabilities access religious materials. In the context of da’wah, AI can broaden the dissemination of knowledge at a fraction of the cost of conventional methods. History shows a similar pattern: the printing press, radio, television, and the internet were all once viewed as threats before ultimately becoming effective tools for spreading religious teachings. The problem is that technological convenience often makes people forget that religion is not merely about information, but also about authority and responsibility.

When AI errs, no one can be held accountable. Modern AI models work by predicting the next word based on patterns learned from billions of data points. They do not understand truth as humans do. Consequently, AI can produce answers that sound highly convincing but are factually incorrect—a phenomenon known in the field as hallucination. These errors might involve incorrect verse numbers, hadith that have no source in any canonical collection, or legal conclusions that contradict scholarly consensus, all delivered with the same confident tone. More dangerously, AI cannot be asked for clarification. There is no teacher, no chain of transmission, and no party that bears moral responsibility for every sentence it produces.

In the Islamic tradition, this issue has been anticipated for centuries through the concept of sanad, or the chain of transmission. Imam Abdullah bin al-Mubarak once stated that sanad is part of the religion; without it, anyone would be free to speak in the name of faith. Imam Muslim even opened his Sahih with a discussion on the importance of sanad before presenting authentic hadith, demonstrating that Islam safeguards not only the content of its teachings but also the path of their transmission. Sanad is not merely a list of teachers; it is a scholarly guarantee system that ensures every link in the chain of knowledge transmission meets the requirements of competence and integrity. It is a mechanism of accountability: from whom knowledge was obtained, how the learning process occurred, and to whom one can turn for clarification when errors arise. In this system, authority is always accompanied by responsibility. An AI persona possesses none of these mechanisms. It has never studied under a teacher, holds no scholarly certification, does not understand the social context in which a fatwa is applied, and cannot be held morally accountable for its statements. The core issue, therefore, is not whether AI can speak, but whether society can distinguish between information and authority.

The Indonesian Ulema Council (MUI) has issued a firm reminder on this matter: AI cannot replace the function of ulama in issuing fatwas or providing tafsir. AI can serve as a tool, but it cannot become a marji’iyyah—a primary reference in religious matters. A fatwa is not merely the result of an information search; it is the fruit of understanding scriptural evidence, reality, the objectives of Sharia, and moral responsibility before God.

This phenomenon is further complicated by the algorithm-driven social media ecosystem. Digital platforms are not designed to select the most accurate content, but rather the content most capable of retaining user attention. Short, emotional, and visually appealing material tends to receive far wider distribution than in-depth scholarly explanations. The MUI has termed this ‘algorithmic religion’—a religious orientation shaped by algorithms rather than authoritative institutions. Data indicates that over 60 per cent of Indonesia’s younger generation now primarily obtain their religious understanding from digital spaces (MUI, 2024). As a result, the public is potentially more likely to encounter ‘ustadz’ favoured by algorithms than scholars who genuinely possess scholarly authority. Authority is gradually shifting from competence to popularity.

The public’s relationship with AI figures also extends beyond mere information consumption. Modern psychology recognises the concept of parasocial relationships—emotional attachments to media figures with whom the audience has no real connection. When an AI persona is designed to appear friendly, empathetic, and perpetually available, some users may develop a very real trust in a figure that never actually existed. This trust can displace bonds with real religious communities—study circles, scholars, and pesantren—which require direct engagement. Simultaneously, the advancement of deepfake technology opens the door to far more serious misuse. A scholar’s face and voice can be counterfeited to deliver a fatwa they never uttered, solicit fraudulent donations, or disseminate deviant teachings. The public’s ability to distinguish between the authentic and the synthetic is now at stake.