Why can't I use ChatGPT and other AI tools to learn Islam?
- Minamuddin Ahmed
- Oct 18, 2023
- 4 min read
Updated: Aug 19
Summary of the Article
This article warns Muslims against relying on AI tools for Islamic knowledge due to their tendency to produce inaccurate information and fabricate references. Through practical examples, the author demonstrates how ChatGPT provided false citations for Islamic rulings. The article emphasizes that Islamic knowledge must come from authentic sources and qualified scholars, as Muslims are religiously obligated to verify information before accepting or sharing it. While AI can serve as a starting point, proper verification through established texts and scholarly consultation is essential.
Religious learning
The knowledge and practice of Islam is the most crucial thing in a Muslim's life. The knowledge acquired must be accurate, clear and backed by authentic Islamic sources.
As Allah says in the Holy Quran 17:36
"And pursue not that of which thou hast no knowledge; for every act of hearing, or of seeing or of (feeling in) the heart will be enquired into (on the Day of Reckoning)."
It is not permissible for anyone to seek or transmit information about Allah, His prophets, His books and His religion in general, without being sure of the authenticity of the content. It is also not acceptable for a Muslim to send and receive information without taking adequate caution to ensure correctness.
Generative AI (GenAI) creates new data based on the large amounts of training data. So, an AI system by itself has no definite mechanism of drawing conclusions and it can sometimes output inaccurate information.
According to ChatGPT's license footnote:
"ChatGPT may produce inaccurate information about people, places, or facts."
Other GenAIs like Claude and Microsoft Copilot also have similar disclaimers.
Furthermore, GenAI has the following major limitations:
Does not understand reality
Does not have personal experience
Is susceptible to bias
Does not have emotions or empathy
Can be easily manipulated
It is very easy for false information about Islam to spread. All Muslims must take extreme caution.
What should we do:
Learn only from authentic sources and qualified scholars
Verify information you receive by crosschecking it
Do not forward information unless you are sure of its correctness
Challenges in Researching Islamic Topics with AI
Recently, I needed to research authentic sources related to the prohibition of listening to deviant preachers. When I prompted ChatGPT, I received a detailed answer with two types of references:
Publicly available sources: References from the Quran and Sunnah were fairly accurate and easy to cross-check using reliable websites.
Copyrighted sources: References from printed books, online fatwa archives, and scholarly lectures often contained inaccurate citations.
For example, ChatGPT quoted a fatwa from Shaykh Ibn Baz and cited "Majmoo' Fatawa Ibn Baaz (3/295)." However, when I checked the actual PDF, I couldn't find the quote.
When questioned about this discrepancy, ChatGPT offered several explanations:
The quote might have been misappropriated or paraphrased
The translation might differ from the original
The quote might exist in a separate fatwa outside the referenced archive
Hallucination in AI
All Generative LLMs are prone to generate incorrect or fabricated information, even when they appear confident. OpenAI models are reported to have a 10% chance of producing incorrect information. ChatGPT head Nick Turley told The Verge, "Until I think we are provably more reliable than a human expert on all domains, not just some domains, I think we're going to continue to advise you to double check your answer."
Why doesn’t AI admit if it does not know?
LLMs are designed to use their large amount of training data and the Internet resources to attempt to answer each question. There have been some reports (as of August 2025) that ChatGPT is replying with ‘I don’t know and I can’t reliably find out” in some cases. This is a positive step. Saying "I don't know" is better than confidently providing plausible-sounding incorrect information. But these honest responses are exceptions and don’t mean that the problem of AI hallucination is reliably fixed.
How to Ensure Accuracy of AI-Provided References
If you do use AI tools for Islamic research, consider these verification strategies:
Verify all references independently - Even when AI provides detailed citations with book titles and page numbers, always check the original source.
Watch for paraphrasing - Determine if the AI has paraphrased the original text and ensure you use such content appropriately.
Refine your prompts - If you don't get the expected results, rephrase your question and clearly explain what you're looking for. Being specific helps AI avoid making incorrect assumptions.
Use multiple research methods - Supplement AI responses with conventional searches through Google or other research tools.
Consult qualified scholars - For critical information, reach out to reputable Islamic scholars or institutions.
Engage with knowledgeable communities - Ask questions in specialized forums and communities focused on Islamic studies.
Think of AI as a hardworking but sometimes unreliable assistant that may not always do its homework properly. For accurate Islamic research:
Be specific in your questions
Ask follow-up questions when needed
Verify information through multiple sources and qualified experts
Can you train ChatGPT to give better answers?
It depends...! While you can correct and educate ChatGPT during your chats, this knowledge remains limited to its conversations with you. This knowledge does not directly cascade and improve ChatGPT's conversations with other users globally- what is called its "global model".
If you want to contribute to the training of ChatGPT's global model, then here are a few options:
Click the thumbs-down ("👎") button on any incorrect or problematic answer.
Provide feedback using the Feedback Form
OpenAI periodically invites community experts to participate in feedback programs, keep an eye on their blog at https://openai.com/blog
Useful links