Dataset Card for gpt_conversations_93k

gpt_conversations_93k is a dataset of more than 93,000 public conversations with ChatGPT retrieved in August 2025. These conversations were publicly available on Google Search after having been shared by the users themselves.

Dataset Structure

JSONL file with 3 fields:

• conversation_id: URL of the ChatGPT conversation
• messages: list of messages in ChatML format
• lang: (main) language of the conversation, automatically annotated with lingua-py

Personal Information

Even though these conversations were shared publicly by the users themselves, the conversations may still contain personal information. By requesting access to the dataset, you explicitly agree that you will use this data for research purposes only and that you will not use this data in any harmful way.

Bias, Risks, and Limitations

Some of the conversations may contain harmful, offensive, misleading, or otherwise inappropriate content generated by users, including language that reflects personal biases, misinformation, or discriminatory viewpoints.

Citation

@misc{enzo_doyen_2025,
    author       = {Enzo Doyen},
    title        = {gpt_conversations_93k (Revision f64796d)},
    year         = 2025,
    url          = {https://huggingface.co/datasets/spidersouris/gpt_conversations_93k},
    doi          = {10.57967/hf/6219},
    publisher    = {Hugging Face}
}

Dataset Card Authors

Enzo Doyen: https://edoyen.com/