Write a PREreview

Large language models can now generate sermons, prayers, doctrinal explanations, and testimonially styled religious speech. Epistemologists have begun to debate whether artificial outputs can count as testimony, while practical-theological studies show that pastors can use AI as a drafting aid without audiences necessarily detecting the assistance. This article asks a more specifically theological question: can words generated by an entity incapable of belief, confession, repentance, commitment, or accountability become Christian witness? It argues that linguistic orthodoxy is insufficient for witnessing agency. An AI system may generate proclamation-content without itself performing the act of proclamation in the full testimonial-theological sense. The article develops the category of Derivative Witness: AI-generated language may enter a genuinely human act of Christian witness when a responsible speaker verifies the content, assents to it, intentionally appropriates it, and accepts answerability for its public use. Extended discussion of speech-act theory, the theology of inspired and derivative speech, and six illustrative cases, from blind reading of AI text to human-authored preaching with AI editing, shows how agency shifts with adoption and responsibility. The article closes with anticipated objections, practical implications for disclosure and formation, and a bounded research agenda. The proposal offers a middle position between treating AI as a preacher and prohibiting all AI assistance.